VBoxNetFlt-linux.c@ 104760

Last change on this file since 104760 was 104046, checked in by vboxsync, 10 months ago
Linux: VBoxNetFlt: Introduce initial support for kernel 6.9, bugref:10630.
Property svn:eol-style set to `native` Property svn:keywords set to `Author Date Id Revision`
File size: 90.5 KB

Line
1	/* $Id: VBoxNetFlt-linux.c 104046 2024-03-25 18:02:43Z vboxsync $ */
2	/** @file
3	* VBoxNetFlt - Network Filter Driver (Host), Linux Specific Code.
4	*/
5
6	/*
7	* Copyright (C) 2006-2023 Oracle and/or its affiliates.
8	*
9	* This file is part of VirtualBox base platform packages, as
10	* available from https://www.virtualbox.org.
11	*
12	* This program is free software; you can redistribute it and/or
13	* modify it under the terms of the GNU General Public License
14	* as published by the Free Software Foundation, in version 3 of the
15	* License.
16	*
17	* This program is distributed in the hope that it will be useful, but
18	* WITHOUT ANY WARRANTY; without even the implied warranty of
19	* MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
20	* General Public License for more details.
21	*
22	* You should have received a copy of the GNU General Public License
23	* along with this program; if not, see <https://www.gnu.org/licenses>.
24	*
25	* The contents of this file may alternatively be used under the terms
26	* of the Common Development and Distribution License Version 1.0
27	* (CDDL), a copy of it is provided in the "COPYING.CDDL" file included
28	* in the VirtualBox distribution, in which case the provisions of the
29	* CDDL are applicable instead of those of the GPL.
30	*
31	* You may elect to license modified versions of this file under the
32	* terms and conditions of either the GPL or the CDDL or both.
33	*
34	* SPDX-License-Identifier: GPL-3.0-only OR CDDL-1.0
35	*/
36
37
38	/*********************************************************************************************************************************
39	* Header Files *
40	*********************************************************************************************************************************/
41	#define LOG_GROUP LOG_GROUP_NET_FLT_DRV
42	#define VBOXNETFLT_LINUX_NO_XMIT_QUEUE
43	#include "the-linux-kernel.h"
44	#include "version-generated.h"
45	#include "revision-generated.h"
46	#include "product-generated.h"
47	#if RTLNX_VER_MIN(2,6,24)
48	# include <linux/nsproxy.h>
49	#endif
50	#if RTLNX_VER_MIN(6,4,10) \|\| RTLNX_RHEL_RANGE(9,4, 9,99)
51	# include <net/gso.h>
52	#endif
53	#include <linux/netdevice.h>
54	#if RTLNX_VER_MAX(2,6,29) \|\| RTLNX_VER_MIN(5,11,0)
55	# include <linux/ethtool.h>
56	#endif
57	#include <linux/etherdevice.h>
58	#include <linux/rtnetlink.h>
59	#include <linux/miscdevice.h>
60	#include <linux/inetdevice.h>
61	#include <linux/in.h>
62	#include <linux/ip.h>
63	#include <linux/if_vlan.h>
64	#if RTLNX_VER_MIN(4,5,0)
65	# include <uapi/linux/pkt_cls.h>
66	#endif
67	#include <net/ipv6.h>
68	#include <net/if_inet6.h>
69	#include <net/addrconf.h>
70
71	#include <VBox/log.h>
72	#include <VBox/err.h>
73	#include <VBox/intnetinline.h>
74	#include <VBox/vmm/pdmnetinline.h>
75	#include <VBox/param.h>
76	#include <VBox/VBoxLnxModInline.h>
77	#include <iprt/alloca.h>
78	#include <iprt/assert.h>
79	#include <iprt/spinlock.h>
80	#include <iprt/semaphore.h>
81	#include <iprt/initterm.h>
82	#include <iprt/process.h>
83	#include <iprt/mem.h>
84	#include <iprt/net.h>
85	#include <iprt/log.h>
86	#include <iprt/mp.h>
87	#include <iprt/mem.h>
88	#include <iprt/time.h>
89
90	#define VBOXNETFLT_OS_SPECFIC 1
91	#include "../VBoxNetFltInternal.h"
92
93	typedef struct VBOXNETFLTNOTIFIER {
94	struct notifier_block Notifier;
95	PVBOXNETFLTINS pThis;
96	} VBOXNETFLTNOTIFIER;
97	typedef struct VBOXNETFLTNOTIFIER *PVBOXNETFLTNOTIFIER;
98
99
100	/*********************************************************************************************************************************
101	* Defined Constants And Macros *
102	*********************************************************************************************************************************/
103	#define VBOX_FLT_NB_TO_INST(pNB) RT_FROM_MEMBER(pNB, VBOXNETFLTINS, u.s.Notifier)
104	#define VBOX_FLT_PT_TO_INST(pPT) RT_FROM_MEMBER(pPT, VBOXNETFLTINS, u.s.PacketType)
105	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
106	# define VBOX_FLT_XT_TO_INST(pXT) RT_FROM_MEMBER(pXT, VBOXNETFLTINS, u.s.XmitTask)
107	#endif
108
109	#if RTLNX_VER_MIN(3,11,0)
110	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) netdev_notifier_info_to_dev(ptr)
111	#else
112	# define VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr) ((struct net_device *)ptr)
113	#endif
114
115	#if RTLNX_VER_MIN(3,5,0)
116	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag))
117	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr)
118	#else
119	# if RTLNX_VER_MIN(3,2,0)
120	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(skb_frag_page(frag), KM_SKB_DATA_SOFTIRQ)
121	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
122	# else
123	# define VBOX_SKB_KMAP_FRAG(frag) kmap_atomic(frag->page, KM_SKB_DATA_SOFTIRQ)
124	# define VBOX_SKB_KUNMAP_FRAG(vaddr) kunmap_atomic(vaddr, KM_SKB_DATA_SOFTIRQ)
125	# endif
126	#endif
127
128	#if RTLNX_VER_MIN(2,6,34)
129	# define VBOX_NETDEV_NAME(dev) netdev_name(dev)
130	#else
131	# define VBOX_NETDEV_NAME(dev) ((dev)->reg_state != NETREG_REGISTERED ? "(unregistered net_device)" : (dev)->name)
132	#endif
133
134	#if RTLNX_VER_MIN(2,6,25)
135	# define VBOX_IPV4_IS_LOOPBACK(addr) ipv4_is_loopback(addr)
136	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ipv4_is_linklocal_169(addr)
137	#else
138	# define VBOX_IPV4_IS_LOOPBACK(addr) ((addr & htonl(IN_CLASSA_NET)) == htonl(0x7f000000))
139	# define VBOX_IPV4_IS_LINKLOCAL_169(addr) ((addr & htonl(IN_CLASSB_NET)) == htonl(0xa9fe0000))
140	#endif
141
142	#if RTLNX_VER_MIN(2,6,22)
143	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb_reset_network_header(skb)
144	# define VBOX_SKB_RESET_MAC_HDR(skb) skb_reset_mac_header(skb)
145	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum_offset
146	#else
147	# define VBOX_SKB_RESET_NETWORK_HDR(skb) skb->nh.raw = skb->data
148	# define VBOX_SKB_RESET_MAC_HDR(skb) skb->mac.raw = skb->data
149	# define VBOX_SKB_CSUM_OFFSET(skb) skb->csum
150	#endif
151
152	#if RTLNX_VER_MIN(2,6,19)
153	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb)
154	#else
155	# define CHECKSUM_PARTIAL CHECKSUM_HW
156	# if RTLNX_VER_MIN(2,6,10)
157	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(skb, 0)
158	# else
159	# if RTLNX_VER_MIN(2,6,7)
160	# define VBOX_SKB_CHECKSUM_HELP(skb) skb_checksum_help(&skb, 0)
161	# else
162	# define VBOX_SKB_CHECKSUM_HELP(skb) (!skb_checksum_help(skb))
163	# endif
164	/* Versions prior 2.6.10 use stats for both bstats and qstats */
165	# define bstats stats
166	# define qstats stats
167	# endif
168	#endif
169
170	#if RTLNX_VER_MIN(6,9,0)
171	# define VBOX_SKB_FRAG_LEN(_pFrag) ((_pFrag)->len)
172	# define VBOX_SKB_FRAG_OFFSET(_pFrag) ((_pFrag)->offset)
173	#elif RTLNX_VER_MIN(5,4,0) \|\| RTLNX_SUSE_MAJ_PREREQ(15, 2)
174	# define VBOX_SKB_FRAG_LEN(_pFrag) ((_pFrag)->bv_len)
175	# define VBOX_SKB_FRAG_OFFSET(_pFrag) ((_pFrag)->bv_offset)
176	#else /* < KERNEL_VERSION(5, 4, 0) */
177	# define VBOX_SKB_FRAG_LEN(_pFrag) ((_pFrag)->size)
178	# define VBOX_SKB_FRAG_OFFSET(_pFrag) ((_pFrag)->page_offset)
179	#endif /* > KERNEL_VERSION(6, 9, 0) */
180
181	#if RTLNX_VER_MIN(3,20,0) \|\| RTLNX_RHEL_RANGE(7,2, 8,0) \|\| RTLNX_RHEL_RANGE(6,8, 7,0)
182	# define VBOX_HAVE_SKB_VLAN
183	#endif
184
185	#ifdef VBOX_HAVE_SKB_VLAN
186	# define vlan_tx_tag_get(skb) skb_vlan_tag_get(skb)
187	# define vlan_tx_tag_present(skb) skb_vlan_tag_present(skb)
188	#endif
189
190	#ifndef NET_IP_ALIGN
191	# define NET_IP_ALIGN 2
192	#endif
193
194	#if 1
195	/** Create scatter / gather segments for fragments. When not used, we will
196	* linearize the socket buffer before creating the internal networking SG. */
197	# define VBOXNETFLT_SG_SUPPORT 1
198	#endif
199
200	#if RTLNX_VER_MIN(2,6,18)
201
202	/** Indicates that the linux kernel may send us GSO frames. */
203	# define VBOXNETFLT_WITH_GSO 1
204
205	/** This enables or disables the transmitting of GSO frame from the internal
206	* network and to the host. */
207	# define VBOXNETFLT_WITH_GSO_XMIT_HOST 1
208
209	# if 0 /** @todo This is currently disable because it causes performance loss of 5-10%. */
210	/** This enables or disables the transmitting of GSO frame from the internal
211	* network and to the wire. */
212	# define VBOXNETFLT_WITH_GSO_XMIT_WIRE 1
213	# endif
214
215	/** This enables or disables the forwarding/flooding of GSO frame from the host
216	* to the internal network. */
217	# define VBOXNETFLT_WITH_GSO_RECV 1
218
219	#endif /* RTLNX_VER_MIN(2,6,18) */
220
221	#if RTLNX_VER_MIN(2,6,29)
222	/** This enables or disables handling of GSO frames coming from the wire (GRO). */
223	# define VBOXNETFLT_WITH_GRO 1
224	#endif
225
226	/*
227	* GRO support was backported to RHEL 5.4
228	*/
229	#if RTLNX_RHEL_MAJ_PREREQ(5, 4)
230	# define VBOXNETFLT_WITH_GRO 1
231	#endif
232
233
234	/*********************************************************************************************************************************
235	* Internal Functions *
236	*********************************************************************************************************************************/
237	static int __init VBoxNetFltLinuxInit(void);
238	static void __exit VBoxNetFltLinuxUnload(void);
239	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf);
240
241
242	/*********************************************************************************************************************************
243	* Global Variables *
244	*********************************************************************************************************************************/
245	/**
246	* The (common) global data.
247	*/
248	static VBOXNETFLTGLOBALS g_VBoxNetFltGlobals;
249
250	module_init(VBoxNetFltLinuxInit);
251	module_exit(VBoxNetFltLinuxUnload);
252
253	MODULE_AUTHOR(VBOX_VENDOR);
254	MODULE_DESCRIPTION(VBOX_PRODUCT " Network Filter Driver");
255	MODULE_LICENSE("GPL");
256	#ifdef MODULE_VERSION
257	MODULE_VERSION(VBOX_VERSION_STRING " r" RT_XSTR(VBOX_SVN_REV) " (" RT_XSTR(INTNETTRUNKIFPORT_VERSION) ")");
258	#endif
259
260
261	#if RTLNX_VER_MAX(2,6,12) && defined(LOG_ENABLED)
262	unsigned dev_get_flags(const struct net_device *dev)
263	{
264	unsigned flags;
265
266	flags = (dev->flags & ~(IFF_PROMISC \|
267	IFF_ALLMULTI \|
268	IFF_RUNNING)) \|
269	(dev->gflags & (IFF_PROMISC \|
270	IFF_ALLMULTI));
271
272	if (netif_running(dev) && netif_carrier_ok(dev))
273	flags \|= IFF_RUNNING;
274
275	return flags;
276	}
277	#endif /* RTLNX_VER_MAX(2,6,12) */
278
279
280	/**
281	* Initialize module.
282	*
283	* @returns appropriate status code.
284	*/
285	static int __init VBoxNetFltLinuxInit(void)
286	{
287	int rc;
288
289	/* Check if modue loading was disabled. */
290	if (!vbox_mod_should_load())
291	return -EINVAL;
292
293	/*
294	* Initialize IPRT.
295	*/
296	rc = RTR0Init(0);
297	if (RT_SUCCESS(rc))
298	{
299	Log(("VBoxNetFltLinuxInit\n"));
300
301	/*
302	* Initialize the globals and connect to the support driver.
303	*
304	* This will call back vboxNetFltOsOpenSupDrv (and maybe vboxNetFltOsCloseSupDrv)
305	* for establishing the connect to the support driver.
306	*/
307	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
308	rc = vboxNetFltInitGlobalsAndIdc(&g_VBoxNetFltGlobals);
309	if (RT_SUCCESS(rc))
310	{
311	LogRel(("VBoxNetFlt: Successfully started.\n"));
312	return 0;
313	}
314
315	LogRel(("VBoxNetFlt: failed to initialize device extension (rc=%d)\n", rc));
316	RTR0Term();
317	}
318	else
319	LogRel(("VBoxNetFlt: failed to initialize IPRT (rc=%d)\n", rc));
320
321	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
322	return -RTErrConvertToErrno(rc);
323	}
324
325
326	/**
327	* Unload the module.
328	*
329	* @todo We have to prevent this if we're busy!
330	*/
331	static void __exit VBoxNetFltLinuxUnload(void)
332	{
333	int rc;
334	Log(("VBoxNetFltLinuxUnload\n"));
335	Assert(vboxNetFltCanUnload(&g_VBoxNetFltGlobals));
336
337	/*
338	* Undo the work done during start (in reverse order).
339	*/
340	rc = vboxNetFltTryDeleteIdcAndGlobals(&g_VBoxNetFltGlobals);
341	AssertRC(rc); NOREF(rc);
342
343	RTR0Term();
344
345	memset(&g_VBoxNetFltGlobals, 0, sizeof(g_VBoxNetFltGlobals));
346
347	Log(("VBoxNetFltLinuxUnload - done\n"));
348	}
349
350
351	/**
352	* We filter traffic from the host to the internal network
353	* before it reaches the NIC driver.
354	*
355	* The current code uses a very ugly hack overriding hard_start_xmit
356	* callback in the device structure, but it has been shown to give us a
357	* performance boost of 60-100% though. Eventually we have to find some
358	* less hacky way of getting this job done.
359	*/
360	#define VBOXNETFLT_WITH_HOST2WIRE_FILTER
361
362	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
363
364	# if RTLNX_VER_MAX(2,6,29)
365
366	typedef struct ethtool_ops OVR_OPSTYPE;
367	# define OVR_OPS ethtool_ops
368	# define OVR_XMIT pfnStartXmit
369
370	# else /* RTLNX_VER_MIN(2,6,29) */
371
372	typedef struct net_device_ops OVR_OPSTYPE;
373	# define OVR_OPS netdev_ops
374	# define OVR_XMIT pOrgOps->ndo_start_xmit
375
376	# endif /* RTLNX_VER_MIN(2,6,29) */
377
378	/**
379	* The overridden net_device_ops of the device we're attached to.
380	*
381	* As there is no net_device_ops structure in pre-2.6.29 kernels we override
382	* ethtool_ops instead along with hard_start_xmit callback in net_device
383	* structure.
384	*
385	* This is a very dirty hack that was created to explore how much we can improve
386	* the host to guest transfers by not CC'ing the NIC. It turns out to be
387	* the only way to filter outgoing packets for devices without TX queue.
388	*/
389	typedef struct VBoxNetDeviceOpsOverride
390	{
391	/** Our overridden ops. */
392	OVR_OPSTYPE Ops;
393	/** Magic word. */
394	uint32_t u32Magic;
395	/** Pointer to the original ops. */
396	OVR_OPSTYPE const *pOrgOps;
397	# if RTLNX_VER_MAX(2,6,29)
398	/** Pointer to the original hard_start_xmit function. */
399	int (pfnStartXmit)(struct sk_buff pSkb, struct net_device *pDev);
400	# endif /* RTLNX_VER_MAX(2,6,29) */
401	/** Pointer to the net filter instance. */
402	PVBOXNETFLTINS pVBoxNetFlt;
403	/** The number of filtered packages. */
404	uint64_t cFiltered;
405	/** The total number of packets */
406	uint64_t cTotal;
407	} VBOXNETDEVICEOPSOVERRIDE, *PVBOXNETDEVICEOPSOVERRIDE;
408	/** VBOXNETDEVICEOPSOVERRIDE::u32Magic value. */
409	#define VBOXNETDEVICEOPSOVERRIDE_MAGIC UINT32_C(0x00c0ffee)
410
411	/**
412	* ndo_start_xmit wrapper that drops packets that shouldn't go to the wire
413	* because they belong on the internal network.
414	*
415	* @returns NETDEV_TX_XXX.
416	* @param pSkb The socket buffer to transmit.
417	* @param pDev The net device.
418	*/
419	static int vboxNetFltLinuxStartXmitFilter(struct sk_buff pSkb, struct net_device pDev)
420	{
421	PVBOXNETDEVICEOPSOVERRIDE pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
422	uint8_t abHdrBuf[sizeof(RTNETETHERHDR) + sizeof(uint32_t) + RTNETIPV4_MIN_LEN];
423	PCRTNETETHERHDR pEtherHdr;
424	PINTNETTRUNKSWPORT pSwitchPort;
425	uint32_t cbHdrs;
426
427
428	/*
429	* Validate the override structure.
430	*
431	* Note! We're racing vboxNetFltLinuxUnhookDev here. If this was supposed
432	* to be production quality code, we would have to be much more
433	* careful here and avoid the race.
434	*/
435	if ( !RT_VALID_PTR(pOverride)
436	\|\| pOverride->u32Magic != VBOXNETDEVICEOPSOVERRIDE_MAGIC
437	# if RTLNX_VER_MIN(2,6,29)
438	\|\| !RT_VALID_PTR(pOverride->pOrgOps)
439	# endif
440	)
441	{
442	printk("vboxNetFltLinuxStartXmitFilter: bad override %p\n", pOverride);
443	dev_kfree_skb(pSkb);
444	return NETDEV_TX_OK;
445	}
446	pOverride->cTotal++;
447
448	/*
449	* Do the filtering base on the default OUI of our virtual NICs
450	*
451	* Note! In a real solution, we would ask the switch whether the
452	* destination MAC is 100% to be on the internal network and then
453	* drop it.
454	*/
455	cbHdrs = skb_headlen(pSkb);
456	cbHdrs = RT_MIN(cbHdrs, sizeof(abHdrBuf));
457	pEtherHdr = (PCRTNETETHERHDR)skb_header_pointer(pSkb, 0, cbHdrs, &abHdrBuf[0]);
458	if ( pEtherHdr
459	&& RT_VALID_PTR(pOverride->pVBoxNetFlt)
460	&& (pSwitchPort = pOverride->pVBoxNetFlt->pSwitchPort) != NULL
461	&& RT_VALID_PTR(pSwitchPort)
462	&& cbHdrs >= 6)
463	{
464	INTNETSWDECISION enmDecision;
465
466	/** @todo consider reference counting, etc. */
467	enmDecision = pSwitchPort->pfnPreRecv(pSwitchPort, pEtherHdr, cbHdrs, INTNETTRUNKDIR_HOST);
468	if (enmDecision == INTNETSWDECISION_INTNET)
469	{
470	dev_kfree_skb(pSkb);
471	pOverride->cFiltered++;
472	return NETDEV_TX_OK;
473	}
474	}
475
476	return pOverride->OVR_XMIT(pSkb, pDev);
477	}
478
479	/**
480	* Hooks the device ndo_start_xmit operation of the device.
481	*
482	* @param pThis The net filter instance.
483	* @param pDev The net device.
484	*/
485	static void vboxNetFltLinuxHookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
486	{
487	PVBOXNETDEVICEOPSOVERRIDE pOverride;
488
489	/* Cancel override if ethtool_ops is missing (host-only case, @bugref{5712}) */
490	if (!RT_VALID_PTR(pDev->OVR_OPS))
491	return;
492	pOverride = RTMemAlloc(sizeof(*pOverride));
493	if (!pOverride)
494	return;
495	pOverride->pOrgOps = pDev->OVR_OPS;
496	pOverride->Ops = *pDev->OVR_OPS;
497	# if RTLNX_VER_MAX(2,6,29)
498	pOverride->pfnStartXmit = pDev->hard_start_xmit;
499	# else /* RTLNX_VER_MIN(2,6,29) */
500	pOverride->Ops.ndo_start_xmit = vboxNetFltLinuxStartXmitFilter;
501	# endif /* RTLNX_VER_MIN(2,6,29) */
502	pOverride->u32Magic = VBOXNETDEVICEOPSOVERRIDE_MAGIC;
503	pOverride->cTotal = 0;
504	pOverride->cFiltered = 0;
505	pOverride->pVBoxNetFlt = pThis;
506
507	RTSpinlockAcquire(pThis->hSpinlock); /* (this isn't necessary, but so what) */
508	ASMAtomicWritePtr((void * volatile *)&pDev->OVR_OPS, pOverride);
509	# if RTLNX_VER_MAX(2,6,29)
510	ASMAtomicXchgPtr((void * volatile *)&pDev->hard_start_xmit, vboxNetFltLinuxStartXmitFilter);
511	# endif /* RTLNX_VER_MAX(2,6,29) */
512	RTSpinlockRelease(pThis->hSpinlock);
513	}
514
515	/**
516	* Undos what vboxNetFltLinuxHookDev did.
517	*
518	* @param pThis The net filter instance.
519	* @param pDev The net device. Can be NULL, in which case
520	* we'll try retrieve it from @a pThis.
521	*/
522	static void vboxNetFltLinuxUnhookDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
523	{
524	PVBOXNETDEVICEOPSOVERRIDE pOverride;
525
526	RTSpinlockAcquire(pThis->hSpinlock);
527	if (!pDev)
528	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
529	if (RT_VALID_PTR(pDev))
530	{
531	pOverride = (PVBOXNETDEVICEOPSOVERRIDE)pDev->OVR_OPS;
532	if ( RT_VALID_PTR(pOverride)
533	&& pOverride->u32Magic == VBOXNETDEVICEOPSOVERRIDE_MAGIC
534	&& RT_VALID_PTR(pOverride->pOrgOps)
535	)
536	{
537	# if RTLNX_VER_MAX(2,6,29)
538	ASMAtomicWritePtr((void * volatile *)&pDev->hard_start_xmit, pOverride->pfnStartXmit);
539	# endif /* RTLNX_VER_MAX(2,6,29) */
540	ASMAtomicWritePtr((void const * volatile *)&pDev->OVR_OPS, pOverride->pOrgOps);
541	ASMAtomicWriteU32(&pOverride->u32Magic, 0);
542	}
543	else
544	pOverride = NULL;
545	}
546	else
547	pOverride = NULL;
548	RTSpinlockRelease(pThis->hSpinlock);
549
550	if (pOverride)
551	{
552	printk("vboxnetflt: %llu out of %llu packets were not sent (directed to host)\n", pOverride->cFiltered, pOverride->cTotal);
553	RTMemFree(pOverride);
554	}
555	}
556
557	#endif /* VBOXNETFLT_WITH_HOST2WIRE_FILTER */
558
559
560	/**
561	* Reads and retains the host interface handle.
562	*
563	* @returns The handle, NULL if detached.
564	* @param pThis
565	*/
566	DECLINLINE(struct net_device *) vboxNetFltLinuxRetainNetDev(PVBOXNETFLTINS pThis)
567	{
568	#if 0
569	struct net_device *pDev = NULL;
570
571	Log(("vboxNetFltLinuxRetainNetDev\n"));
572	/*
573	* Be careful here to avoid problems racing the detached callback.
574	*/
575	RTSpinlockAcquire(pThis->hSpinlock);
576	if (!ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost))
577	{
578	pDev = (struct net_device )ASMAtomicUoReadPtr((void volatile *)&pThis->u.s.pDev);
579	if (pDev)
580	{
581	dev_hold(pDev);
582	Log(("vboxNetFltLinuxRetainNetDev: Device %p(%s) retained. ref=%d\n",
583	pDev, pDev->name,
584	#if RTLNX_VER_MIN(2,6,37)
585	netdev_refcnt_read(pDev)
586	#else
587	atomic_read(&pDev->refcnt)
588	#endif
589	));
590	}
591	}
592	RTSpinlockRelease(pThis->hSpinlock);
593
594	Log(("vboxNetFltLinuxRetainNetDev - done\n"));
595	return pDev;
596	#else
597	return ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
598	#endif
599	}
600
601
602	/**
603	* Release the host interface handle previously retained
604	* by vboxNetFltLinuxRetainNetDev.
605	*
606	* @param pThis The instance.
607	* @param pDev The vboxNetFltLinuxRetainNetDev
608	* return value, NULL is fine.
609	*/
610	DECLINLINE(void) vboxNetFltLinuxReleaseNetDev(PVBOXNETFLTINS pThis, struct net_device *pDev)
611	{
612	#if 0
613	Log(("vboxNetFltLinuxReleaseNetDev\n"));
614	NOREF(pThis);
615	if (pDev)
616	{
617	dev_put(pDev);
618	Log(("vboxNetFltLinuxReleaseNetDev: Device %p(%s) released. ref=%d\n",
619	pDev, pDev->name,
620	#if RTLNX_VER_MIN(2,6,37)
621	netdev_refcnt_read(pDev)
622	#else
623	atomic_read(&pDev->refcnt)
624	#endif
625	));
626	}
627	Log(("vboxNetFltLinuxReleaseNetDev - done\n"));
628	#endif
629	}
630
631	#define VBOXNETFLT_CB_TAG(skb) (0xA1C90000 \| (skb->dev->ifindex & 0xFFFF))
632	#define VBOXNETFLT_SKB_TAG(skb) ((uint32_t)&((skb)->cb[sizeof((skb)->cb)-sizeof(uint32_t)]))
633
634	/**
635	* Checks whether this is an mbuf created by vboxNetFltLinuxMBufFromSG,
636	* i.e. a buffer which we're pushing and should be ignored by the filter callbacks.
637	*
638	* @returns true / false accordingly.
639	* @param pBuf The sk_buff.
640	*/
641	DECLINLINE(bool) vboxNetFltLinuxSkBufIsOur(struct sk_buff *pBuf)
642	{
643	return VBOXNETFLT_SKB_TAG(pBuf) == VBOXNETFLT_CB_TAG(pBuf);
644	}
645
646
647	/**
648	* Checks whether this SG list contains a GSO packet.
649	*
650	* @returns true / false accordingly.
651	* @param pSG The (scatter/)gather list.
652	*/
653	DECLINLINE(bool) vboxNetFltLinuxIsGso(PINTNETSG pSG)
654	{
655	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
656	return !((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type == PDMNETWORKGSOTYPE_INVALID);
657	#else /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
658	return false;
659	#endif /* !VBOXNETFLT_WITH_GSO_XMIT_WIRE && !VBOXNETFLT_WITH_GSO_XMIT_HOST */
660	}
661
662
663	/**
664	* Find out the frame size (of a single segment in case of GSO frames).
665	*
666	* @returns the frame size.
667	* @param pSG The (scatter/)gather list.
668	*/
669	DECLINLINE(uint32_t) vboxNetFltLinuxFrameSize(PINTNETSG pSG)
670	{
671	uint16_t u16Type = 0;
672	uint32_t cbVlanTag = 0;
673	if (pSG->aSegs[0].cb >= sizeof(RTNETETHERHDR))
674	u16Type = RT_BE2H_U16(((PCRTNETETHERHDR)pSG->aSegs[0].pv)->EtherType);
675	else if (pSG->cbTotal >= sizeof(RTNETETHERHDR))
676	{
677	uint32_t off = RT_UOFFSETOF(RTNETETHERHDR, EtherType);
678	uint32_t i;
679	for (i = 0; i < pSG->cSegsUsed; ++i)
680	{
681	if (off <= pSG->aSegs[i].cb)
682	{
683	if (off + sizeof(uint16_t) <= pSG->aSegs[i].cb)
684	u16Type = RT_BE2H_U16((uint16_t )((uintptr_t)pSG->aSegs[i].pv + off));
685	else if (i + 1 < pSG->cSegsUsed)
686	u16Type = RT_BE2H_U16( ((uint16_t)( ((uint8_t *)pSG->aSegs[i].pv)[off] ) << 8)
687	+ (uint8_t )pSG->aSegs[i + 1].pv); /* ASSUMES no empty segments! */
688	/* else: frame is too short. */
689	break;
690	}
691	off -= pSG->aSegs[i].cb;
692	}
693	}
694	if (u16Type == RTNET_ETHERTYPE_VLAN)
695	cbVlanTag = 4;
696	return (vboxNetFltLinuxIsGso(pSG) ? (uint32_t)pSG->GsoCtx.cbMaxSeg + pSG->GsoCtx.cbHdrsTotal : pSG->cbTotal) - cbVlanTag;
697	}
698
699
700	/**
701	* Internal worker that create a linux sk_buff for a
702	* (scatter/)gather list.
703	*
704	* @returns Pointer to the sk_buff.
705	* @param pThis The instance.
706	* @param pSG The (scatter/)gather list.
707	* @param fDstWire Set if the destination is the wire.
708	*/
709	static struct sk_buff *vboxNetFltLinuxSkBufFromSG(PVBOXNETFLTINS pThis, PINTNETSG pSG, bool fDstWire)
710	{
711	struct sk_buff *pPkt;
712	struct net_device *pDev;
713	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
714	unsigned fGsoType = 0;
715	#endif
716
717	if (pSG->cbTotal == 0)
718	{
719	LogRel(("VBoxNetFlt: Dropped empty packet coming from internal network.\n"));
720	return NULL;
721	}
722	Log5(("VBoxNetFlt: Packet to %s of %d bytes (frame=%d).\n", fDstWire?"wire":"host", pSG->cbTotal, vboxNetFltLinuxFrameSize(pSG)));
723	if (fDstWire && (vboxNetFltLinuxFrameSize(pSG) > ASMAtomicReadU32(&pThis->u.s.cbMtu) + 14))
724	{
725	static bool s_fOnce = true;
726	if (s_fOnce)
727	{
728	s_fOnce = false;
729	printk("VBoxNetFlt: Dropped over-sized packet (%d bytes) coming from internal network.\n", vboxNetFltLinuxFrameSize(pSG));
730	}
731	return NULL;
732	}
733
734	/** @todo We should use fragments mapping the SG buffers with large packets.
735	* 256 bytes seems to be the a threshold used a lot for this. It
736	* requires some nasty work on the intnet side though... */
737	/*
738	* Allocate a packet and copy over the data.
739	*/
740	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
741	pPkt = dev_alloc_skb(pSG->cbTotal + NET_IP_ALIGN);
742	if (RT_UNLIKELY(!pPkt))
743	{
744	Log(("vboxNetFltLinuxSkBufFromSG: Failed to allocate sk_buff(%u).\n", pSG->cbTotal));
745	pSG->pvUserData = NULL;
746	return NULL;
747	}
748	pPkt->dev = pDev;
749	pPkt->ip_summed = CHECKSUM_NONE;
750
751	/* Align IP header on 16-byte boundary: 2 + 14 (ethernet hdr size). */
752	skb_reserve(pPkt, NET_IP_ALIGN);
753
754	/* Copy the segments. */
755	skb_put(pPkt, pSG->cbTotal);
756	IntNetSgRead(pSG, pPkt->data);
757
758	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
759	/*
760	* Setup GSO if used by this packet.
761	*/
762	switch ((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type)
763	{
764	default:
765	AssertMsgFailed(("%u (%s)\n", pSG->GsoCtx.u8Type, PDMNetGsoTypeName((PDMNETWORKGSOTYPE)pSG->GsoCtx.u8Type) ));
766	RT_FALL_THRU();
767	case PDMNETWORKGSOTYPE_INVALID:
768	fGsoType = 0;
769	break;
770	case PDMNETWORKGSOTYPE_IPV4_TCP:
771	fGsoType = SKB_GSO_TCPV4;
772	break;
773	case PDMNETWORKGSOTYPE_IPV6_TCP:
774	fGsoType = SKB_GSO_TCPV6;
775	break;
776	}
777	if (fGsoType)
778	{
779	struct skb_shared_info *pShInfo = skb_shinfo(pPkt);
780
781	pShInfo->gso_type = fGsoType \| SKB_GSO_DODGY;
782	pShInfo->gso_size = pSG->GsoCtx.cbMaxSeg;
783	pShInfo->gso_segs = PDMNetGsoCalcSegmentCount(&pSG->GsoCtx, pSG->cbTotal);
784
785	/*
786	* We need to set checksum fields even if the packet goes to the host
787	* directly as it may be immediately forwarded by IP layer @bugref{5020}.
788	*/
789	Assert(skb_headlen(pPkt) >= pSG->GsoCtx.cbHdrsTotal);
790	pPkt->ip_summed = CHECKSUM_PARTIAL;
791	# if RTLNX_VER_MIN(2,6,22)
792	pPkt->csum_start = skb_headroom(pPkt) + pSG->GsoCtx.offHdr2;
793	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
794	pPkt->csum_offset = RT_UOFFSETOF(RTNETTCP, th_sum);
795	else
796	pPkt->csum_offset = RT_UOFFSETOF(RTNETUDP, uh_sum);
797	# else
798	pPkt->h.raw = pPkt->data + pSG->GsoCtx.offHdr2;
799	if (fGsoType & (SKB_GSO_TCPV4 \| SKB_GSO_TCPV6))
800	pPkt->csum = RT_UOFFSETOF(RTNETTCP, th_sum);
801	else
802	pPkt->csum = RT_UOFFSETOF(RTNETUDP, uh_sum);
803	# endif
804	if (!fDstWire)
805	PDMNetGsoPrepForDirectUse(&pSG->GsoCtx, pPkt->data, pSG->cbTotal, PDMNETCSUMTYPE_PSEUDO);
806	}
807	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
808
809	/*
810	* Finish up the socket buffer.
811	*/
812	pPkt->protocol = eth_type_trans(pPkt, pDev);
813	if (fDstWire)
814	{
815	VBOX_SKB_RESET_NETWORK_HDR(pPkt);
816
817	/* Restore ethernet header back. */
818	skb_push(pPkt, ETH_HLEN); /** @todo VLAN: +4 if VLAN? */
819	VBOX_SKB_RESET_MAC_HDR(pPkt);
820	}
821	VBOXNETFLT_SKB_TAG(pPkt) = VBOXNETFLT_CB_TAG(pPkt);
822
823	return pPkt;
824	}
825
826
827	/**
828	* Return the offset where to start checksum computation from.
829	*
830	* @returns the offset relative to pBuf->data.
831	* @param pBuf The socket buffer.
832	*/
833	DECLINLINE(unsigned) vboxNetFltLinuxGetChecksumStartOffset(struct sk_buff *pBuf)
834	{
835	#if RTLNX_VER_MIN(2,6,38)
836	return skb_checksum_start_offset(pBuf);
837	#elif RTLNX_VER_MIN(2,6,22)
838	return pBuf->csum_start - skb_headroom(pBuf);
839	#else
840	unsigned char *pTransportHdr = pBuf->h.raw;
841	# if RTLNX_VER_MAX(2,6,19)
842	/*
843	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
844	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
845	* header length from the header itself and reconstruct 'h' pointer
846	* to TCP (or whatever) header.
847	*/
848	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
849	pTransportHdr = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
850	# endif
851	return pTransportHdr - pBuf->data;
852	#endif
853	}
854
855
856	/**
857	* Initializes a SG list from an sk_buff.
858	*
859	* @param pThis The instance.
860	* @param pBuf The sk_buff.
861	* @param pSG The SG.
862	* @param cbExtra The number of bytes of extra space allocated immediately after the SG.
863	* @param cSegs The number of segments allocated for the SG.
864	* This should match the number in the mbuf exactly!
865	* @param fSrc The source of the frame.
866	* @param pGsoCtx Pointer to the GSO context if it's a GSO
867	* internal network frame. NULL if regular frame.
868	*/
869	static void vboxNetFltLinuxSkBufToSG(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, PINTNETSG pSG,
870	unsigned cbExtra, unsigned cSegs, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
871	{
872	int i;
873	NOREF(pThis);
874
875	#ifndef VBOXNETFLT_SG_SUPPORT
876	Assert(!skb_shinfo(pBuf)->frag_list);
877	#else /* VBOXNETFLT_SG_SUPPORT */
878	uint8_t pExtra = (uint8_t )&pSG->aSegs[cSegs];
879	unsigned cbConsumed = 0;
880	unsigned cbProduced = 0;
881
882	# if RTLNX_VER_MIN(2,6,27)
883	/* Restore VLAN tag stripped by host hardware */
884	if (vlan_tx_tag_present(pBuf))
885	{
886	uint8_t *pMac = pBuf->data;
887	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )pExtra;
888	Assert(ETH_ALEN * 2 + VLAN_HLEN <= cbExtra);
889	memmove(pVHdr, pMac, ETH_ALEN * 2);
890	/* Consume whole Ethernet header: 2 addresses + EtherType (see @bugref{8599}) */
891	cbConsumed += ETH_ALEN * 2 + sizeof(uint16_t);
892	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
893	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
894	pVHdr->h_vlan_encapsulated_proto = (uint16_t)(pMac + ETH_ALEN * 2);
895	cbProduced += VLAN_ETH_HLEN;
896	}
897	# endif /* RTLNX_VER_MIN(2,6,27) */
898
899	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
900	{
901	unsigned uCsumStartOffset = vboxNetFltLinuxGetChecksumStartOffset(pBuf);
902	unsigned uCsumStoreOffset = uCsumStartOffset + VBOX_SKB_CSUM_OFFSET(pBuf) - cbConsumed;
903	Log3(("cbConsumed=%u cbProduced=%u uCsumStartOffset=%u uCsumStoreOffset=%u\n",
904	cbConsumed, cbProduced, uCsumStartOffset, uCsumStoreOffset));
905	Assert(cbProduced + uCsumStoreOffset + sizeof(uint16_t) <= cbExtra);
906	/*
907	* We assume that the checksum is stored at the very end of the transport header
908	* so we will have all headers in a single fragment. If our assumption is wrong
909	* we may see suboptimal performance.
910	*/
911	memmove(pExtra + cbProduced,
912	pBuf->data + cbConsumed,
913	uCsumStoreOffset);
914	unsigned uChecksum = skb_checksum(pBuf, uCsumStartOffset, pBuf->len - uCsumStartOffset, 0);
915	(uint16_t)(pExtra + cbProduced + uCsumStoreOffset) = csum_fold(uChecksum);
916	cbProduced += uCsumStoreOffset + sizeof(uint16_t);
917	cbConsumed += uCsumStoreOffset + sizeof(uint16_t);
918	}
919	#endif /* VBOXNETFLT_SG_SUPPORT */
920
921	if (!pGsoCtx)
922	IntNetSgInitTempSegs(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/);
923	else
924	IntNetSgInitTempSegsGso(pSG, pBuf->len + cbProduced - cbConsumed, cSegs, 0 /cSegsUsed/, pGsoCtx);
925
926	int iSeg = 0;
927	#ifdef VBOXNETFLT_SG_SUPPORT
928	if (cbProduced)
929	{
930	pSG->aSegs[iSeg].cb = cbProduced;
931	pSG->aSegs[iSeg].pv = pExtra;
932	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
933	}
934	pSG->aSegs[iSeg].cb = skb_headlen(pBuf) - cbConsumed;
935	pSG->aSegs[iSeg].pv = pBuf->data + cbConsumed;
936	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
937	Assert(iSeg <= pSG->cSegsAlloc);
938
939	# ifdef LOG_ENABLED
940	if (pBuf->data_len)
941	Log6((" kmap_atomic:"));
942	# endif /* LOG_ENABLED */
943	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
944	{
945	skb_frag_t *pFrag = &skb_shinfo(pBuf)->frags[i];
946	pSG->aSegs[iSeg].cb = VBOX_SKB_FRAG_LEN(pFrag);
947	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + VBOX_SKB_FRAG_OFFSET(pFrag);
948	Log6((" %p", pSG->aSegs[iSeg].pv));
949	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
950	Assert(iSeg <= pSG->cSegsAlloc);
951	}
952	struct sk_buff *pFragBuf;
953	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
954	{
955	pSG->aSegs[iSeg].cb = skb_headlen(pFragBuf);
956	pSG->aSegs[iSeg].pv = pFragBuf->data;
957	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
958	Assert(iSeg <= pSG->cSegsAlloc);
959	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
960	{
961	skb_frag_t *pFrag = &skb_shinfo(pFragBuf)->frags[i];
962	pSG->aSegs[iSeg].cb = VBOX_SKB_FRAG_LEN(pFrag);
963	pSG->aSegs[iSeg].pv = VBOX_SKB_KMAP_FRAG(pFrag) + VBOX_SKB_FRAG_OFFSET(pFrag);
964	Log6((" %p", pSG->aSegs[iSeg].pv));
965	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
966	Assert(iSeg <= pSG->cSegsAlloc);
967	}
968	}
969	# ifdef LOG_ENABLED
970	if (pBuf->data_len)
971	Log6(("\n"));
972	# endif /* LOG_ENABLED */
973	#else
974	pSG->aSegs[iSeg].cb = pBuf->len;
975	pSG->aSegs[iSeg].pv = pBuf->data;
976	pSG->aSegs[iSeg++].Phys = NIL_RTHCPHYS;
977	#endif
978
979	pSG->cSegsUsed = iSeg;
980
981	#if 0
982	if (cbProduced)
983	{
984	LogRel(("vboxNetFltLinuxSkBufToSG: original packet dump:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
985	LogRel(("vboxNetFltLinuxSkBufToSG: cbConsumed=%u cbProduced=%u cbExtra=%u\n", cbConsumed, cbProduced, cbExtra));
986	uint32_t offset = 0;
987	for (i = 0; i < pSG->cSegsUsed; ++i)
988	{
989	LogRel(("vboxNetFltLinuxSkBufToSG: seg#%d (%d bytes, starting at 0x%x):\n%.*Rhxd\n",
990	i, pSG->aSegs[i].cb, offset, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
991	offset += pSG->aSegs[i].cb;
992	}
993	}
994	#endif
995
996	#ifdef PADD_RUNT_FRAMES_FROM_HOST
997	/*
998	* Add a trailer if the frame is too small.
999	*
1000	* Since we're getting to the packet before it is framed, it has not
1001	* yet been padded. The current solution is to add a segment pointing
1002	* to a buffer containing all zeros and pray that works for all frames...
1003	*/
1004	if (pSG->cbTotal < 60 && (fSrc & INTNETTRUNKDIR_HOST))
1005	{
1006	Assert(pBuf->data_len == 0); /* Packets with fragments are never small! */
1007	static uint8_t const s_abZero[128] = {0};
1008
1009	AssertReturnVoid(iSeg < cSegs);
1010
1011	pSG->aSegs[iSeg].Phys = NIL_RTHCPHYS;
1012	pSG->aSegs[iSeg].pv = (void *)&s_abZero[0];
1013	pSG->aSegs[iSeg++].cb = 60 - pSG->cbTotal;
1014	pSG->cbTotal = 60;
1015	pSG->cSegsUsed++;
1016	Assert(iSeg <= pSG->cSegsAlloc)
1017	}
1018	#endif
1019
1020	Log6(("vboxNetFltLinuxSkBufToSG: allocated=%d, segments=%d frags=%d next=%p frag_list=%p pkt_type=%x fSrc=%x\n",
1021	pSG->cSegsAlloc, pSG->cSegsUsed, skb_shinfo(pBuf)->nr_frags, pBuf->next, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, fSrc));
1022	for (i = 0; i < pSG->cSegsUsed; i++)
1023	Log6(("vboxNetFltLinuxSkBufToSG: #%d: cb=%d pv=%p\n",
1024	i, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1025	}
1026
1027	/**
1028	* Packet handler; not really documented - figure it out yourself.
1029	*
1030	* @returns 0 or EJUSTRETURN - this is probably copy & pastry and thus wrong.
1031	*/
1032	#if RTLNX_VER_MIN(2,6,14)
1033	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1034	struct net_device *pSkbDev,
1035	struct packet_type *pPacketType,
1036	struct net_device *pOrigDev)
1037	#else
1038	static int vboxNetFltLinuxPacketHandler(struct sk_buff *pBuf,
1039	struct net_device *pSkbDev,
1040	struct packet_type *pPacketType)
1041	#endif
1042	{
1043	PVBOXNETFLTINS pThis;
1044	struct net_device *pDev;
1045	LogFlow(("vboxNetFltLinuxPacketHandler: pBuf=%p pSkbDev=%p pPacketType=%p\n",
1046	pBuf, pSkbDev, pPacketType));
1047	#if RTLNX_VER_MIN(2,6,18)
1048	Log3(("vboxNetFltLinuxPacketHandler: skb len=%u data_len=%u truesize=%u next=%p nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1049	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size, skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1050	# if RTLNX_VER_MIN(2,6,22)
1051	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1052	# endif
1053	#else
1054	Log3(("vboxNetFltLinuxPacketHandler: skb len=%u data_len=%u truesize=%u next=%p nr_frags=%u tso_size=%u tso_seqs=%u frag_list=%p pkt_type=%x\n",
1055	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->tso_size, skb_shinfo(pBuf)->tso_segs, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1056	#endif
1057	/*
1058	* Drop it immediately?
1059	*/
1060	if (!pBuf)
1061	return 0;
1062
1063	if (pBuf->pkt_type == PACKET_LOOPBACK)
1064	{
1065	/*
1066	* We are not interested in loopbacked packets as they will always have
1067	* another copy going to the wire.
1068	*/
1069	Log2(("vboxNetFltLinuxPacketHandler: dropped loopback packet (cb=%u)\n", pBuf->len));
1070	dev_kfree_skb(pBuf); /* We must 'consume' all packets we get (@bugref{6539})! */
1071	return 0;
1072	}
1073
1074	pThis = VBOX_FLT_PT_TO_INST(pPacketType);
1075	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1076	if (pDev != pSkbDev)
1077	{
1078	Log(("vboxNetFltLinuxPacketHandler: Devices do not match, pThis may be wrong! pThis=%p\n", pThis));
1079	kfree_skb(pBuf); /* This is a failure, so we use kfree_skb instead of dev_kfree_skb. */
1080	return 0;
1081	}
1082
1083	Log6(("vboxNetFltLinuxPacketHandler: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
1084	if (vboxNetFltLinuxSkBufIsOur(pBuf))
1085	{
1086	Log2(("vboxNetFltLinuxPacketHandler: got our own sk_buff, drop it.\n"));
1087	dev_kfree_skb(pBuf);
1088	return 0;
1089	}
1090
1091	#ifndef VBOXNETFLT_SG_SUPPORT
1092	{
1093	/*
1094	* Get rid of fragmented packets, they cause too much trouble.
1095	*/
1096	unsigned int uMacLen = pBuf->mac_len;
1097	struct sk_buff *pCopy = skb_copy(pBuf, GFP_ATOMIC);
1098	dev_kfree_skb(pBuf);
1099	if (!pCopy)
1100	{
1101	LogRel(("VBoxNetFlt: Failed to allocate packet buffer, dropping the packet.\n"));
1102	return 0;
1103	}
1104	pBuf = pCopy;
1105	/* Somehow skb_copy ignores mac_len */
1106	pBuf->mac_len = uMacLen;
1107	# if RTLNX_VER_MIN(2,6,27)
1108	/* Restore VLAN tag stripped by host hardware */
1109	if (vlan_tx_tag_present(pBuf) && skb_headroom(pBuf) >= VLAN_ETH_HLEN)
1110	{
1111	uint8_t pMac = (uint8_t)skb_mac_header(pBuf);
1112	struct vlan_ethhdr pVHdr = (struct vlan_ethhdr )(pMac - VLAN_HLEN);
1113	memmove(pVHdr, pMac, ETH_ALEN * 2);
1114	pVHdr->h_vlan_proto = RT_H2N_U16(ETH_P_8021Q);
1115	pVHdr->h_vlan_TCI = RT_H2N_U16(vlan_tx_tag_get(pBuf));
1116	pBuf->mac_header -= VLAN_HLEN;
1117	pBuf->mac_len += VLAN_HLEN;
1118	}
1119	# endif /* RTLNX_VER_MIN(2,6,27) */
1120
1121	# if RTLNX_VER_MIN(2,6,18)
1122	Log3(("vboxNetFltLinuxPacketHandler: skb copy len=%u data_len=%u truesize=%u next=%p nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1123	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size, skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1124	# if RTLNX_VER_MIN(2,6,22)
1125	Log6(("vboxNetFltLinuxPacketHandler: packet dump follows:\n%.*Rhxd\n", pBuf->len-pBuf->data_len, skb_mac_header(pBuf)));
1126	# endif /* RTLNX_VER_MIN(2,6,22) */
1127	# else /* RTLNX_VER_MAX(2,6,18) */
1128	Log3(("vboxNetFltLinuxPacketHandler: skb copy len=%u data_len=%u truesize=%u next=%p nr_frags=%u tso_size=%u tso_seqs=%u frag_list=%p pkt_type=%x\n",
1129	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next, skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->tso_size, skb_shinfo(pBuf)->tso_segs, skb_shinfo(pBuf)->frag_list, pBuf->pkt_type));
1130	# endif /* RTLNX_VER_MAX(2,6,18) */
1131	}
1132	#endif /* !VBOXNETFLT_SG_SUPPORT */
1133
1134	#ifdef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1135	/* Forward it to the internal network. */
1136	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1137	#else /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1138	/* Add the packet to transmit queue and schedule the bottom half. */
1139	skb_queue_tail(&pThis->u.s.XmitQueue, pBuf);
1140	schedule_work(&pThis->u.s.XmitTask);
1141	Log6(("vboxNetFltLinuxPacketHandler: scheduled work %p for sk_buff %p\n",
1142	&pThis->u.s.XmitTask, pBuf));
1143	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1144
1145	/* It does not really matter what we return, it is ignored by the kernel. */
1146	return 0;
1147	}
1148
1149	/**
1150	* Calculate the number of INTNETSEG segments the socket buffer will need.
1151	*
1152	* @returns Segment count.
1153	* @param pBuf The socket buffer.
1154	* @param pcbTemp Where to store the number of bytes of the part
1155	* of the socket buffer that will be copied to
1156	* a temporary storage.
1157	*/
1158	DECLINLINE(unsigned) vboxNetFltLinuxCalcSGSegments(struct sk_buff pBuf, unsigned pcbTemp)
1159	{
1160	*pcbTemp = 0;
1161	#ifdef VBOXNETFLT_SG_SUPPORT
1162	unsigned cSegs = 1 + skb_shinfo(pBuf)->nr_frags;
1163	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1164	{
1165	*pcbTemp = vboxNetFltLinuxGetChecksumStartOffset(pBuf) + VBOX_SKB_CSUM_OFFSET(pBuf) + sizeof(uint16_t);
1166	}
1167	# if RTLNX_VER_MIN(2,6,27)
1168	if (vlan_tx_tag_present(pBuf))
1169	{
1170	if (*pcbTemp)
1171	*pcbTemp += VLAN_HLEN;
1172	else
1173	*pcbTemp = VLAN_ETH_HLEN;
1174	}
1175	# endif /* RTLNX_VER_MIN(2,6,27) */
1176	if (*pcbTemp)
1177	++cSegs;
1178	struct sk_buff *pFrag;
1179	for (pFrag = skb_shinfo(pBuf)->frag_list; pFrag; pFrag = pFrag->next)
1180	{
1181	Log6(("vboxNetFltLinuxCalcSGSegments: frag=%p len=%d data_len=%d frags=%d frag_list=%p next=%p\n",
1182	pFrag, pFrag->len, pFrag->data_len, skb_shinfo(pFrag)->nr_frags, skb_shinfo(pFrag)->frag_list, pFrag->next));
1183	cSegs += 1 + skb_shinfo(pFrag)->nr_frags;
1184	}
1185	#else
1186	unsigned cSegs = 1;
1187	#endif
1188	#ifdef PADD_RUNT_FRAMES_FROM_HOST
1189	/* vboxNetFltLinuxSkBufToSG adds a padding segment if it's a runt. */
1190	if (pBuf->len < 60)
1191	cSegs++;
1192	#endif
1193	return cSegs;
1194	}
1195
1196
1197	/**
1198	* Destroy the intnet scatter / gather buffer created by
1199	* vboxNetFltLinuxSkBufToSG.
1200	*
1201	* @param pSG The (scatter/)gather list.
1202	* @param pBuf The original socket buffer that was used to create
1203	* the scatter/gather list.
1204	*/
1205	static void vboxNetFltLinuxDestroySG(PINTNETSG pSG, struct sk_buff *pBuf)
1206	{
1207	#ifdef VBOXNETFLT_SG_SUPPORT
1208	int i, iSeg = 1; /* Skip non-paged part of SKB */
1209	/* Check if the extra buffer behind SG structure was used for modified packet header */
1210	if (pBuf->data != pSG->aSegs[0].pv)
1211	++iSeg; /* Skip it as well */
1212	# ifdef LOG_ENABLED
1213	if (pBuf->data_len)
1214	Log6(("kunmap_atomic:"));
1215	# endif /* LOG_ENABLED */
1216	/* iSeg now points to the first mapped fragment if there are any */
1217	for (i = 0; i < skb_shinfo(pBuf)->nr_frags; i++)
1218	{
1219	Log6((" %p", pSG->aSegs[iSeg].pv));
1220	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1221	}
1222	struct sk_buff *pFragBuf;
1223	for (pFragBuf = skb_shinfo(pBuf)->frag_list; pFragBuf; pFragBuf = pFragBuf->next)
1224	{
1225	++iSeg; /* Non-fragment (unmapped) portion of chained SKB */
1226	for (i = 0; i < skb_shinfo(pFragBuf)->nr_frags; i++)
1227	{
1228	Log6((" %p", pSG->aSegs[iSeg].pv));
1229	VBOX_SKB_KUNMAP_FRAG(pSG->aSegs[iSeg++].pv);
1230	}
1231	}
1232	# ifdef LOG_ENABLED
1233	if (pBuf->data_len)
1234	Log6(("\n"));
1235	# endif /* LOG_ENABLED */
1236	#endif
1237	NOREF(pSG);
1238	}
1239
1240	#ifdef LOG_ENABLED
1241	/**
1242	* Logging helper.
1243	*/
1244	static void vboxNetFltDumpPacket(PINTNETSG pSG, bool fEgress, const char *pszWhere, int iIncrement)
1245	{
1246	int i, offSeg;
1247	uint8_t pInt, pExt;
1248	static int iPacketNo = 1;
1249	iPacketNo += iIncrement;
1250	if (fEgress)
1251	{
1252	pExt = pSG->aSegs[0].pv;
1253	pInt = pExt + 6;
1254	}
1255	else
1256	{
1257	pInt = pSG->aSegs[0].pv;
1258	pExt = pInt + 6;
1259	}
1260	Log(("VBoxNetFlt: (int)%02x:%02x:%02x:%02x:%02x:%02x"
1261	" %s (%s)%02x:%02x:%02x:%02x:%02x:%02x (%u bytes) packet #%u\n",
1262	pInt[0], pInt[1], pInt[2], pInt[3], pInt[4], pInt[5],
1263	fEgress ? "-->" : "<--", pszWhere,
1264	pExt[0], pExt[1], pExt[2], pExt[3], pExt[4], pExt[5],
1265	pSG->cbTotal, iPacketNo));
1266	if (pSG->cSegsUsed == 1)
1267	{
1268	Log4(("%.*Rhxd\n", pSG->aSegs[0].cb, pSG->aSegs[0].pv));
1269	}
1270	else
1271	{
1272	for (i = 0, offSeg = 0; i < pSG->cSegsUsed; i++)
1273	{
1274	Log4(("-- segment %d at 0x%x (%d bytes)\n --\n%.*Rhxd\n",
1275	i, offSeg, pSG->aSegs[i].cb, pSG->aSegs[i].cb, pSG->aSegs[i].pv));
1276	offSeg += pSG->aSegs[i].cb;
1277	}
1278	}
1279	}
1280	#else
1281	# define vboxNetFltDumpPacket(a, b, c, d) do {} while (0)
1282	#endif
1283
1284	#ifdef VBOXNETFLT_WITH_GSO_RECV
1285
1286	/**
1287	* Worker for vboxNetFltLinuxForwardToIntNet that checks if we can forwards a
1288	* GSO socket buffer without having to segment it.
1289	*
1290	* @returns true on success, false if needs segmenting.
1291	* @param pThis The net filter instance.
1292	* @param pSkb The GSO socket buffer.
1293	* @param fSrc The source.
1294	* @param pGsoCtx Where to return the GSO context on success.
1295	*/
1296	static bool vboxNetFltLinuxCanForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc,
1297	PPDMNETWORKGSO pGsoCtx)
1298	{
1299	PDMNETWORKGSOTYPE enmGsoType;
1300	uint16_t uEtherType;
1301	unsigned int cbTransport;
1302	unsigned int offTransport;
1303	unsigned int cbTransportHdr;
1304	unsigned uProtocol;
1305	union
1306	{
1307	RTNETIPV4 IPv4;
1308	RTNETIPV6 IPv6;
1309	RTNETTCP Tcp;
1310	uint8_t ab[40];
1311	uint16_t au16[40/2];
1312	uint32_t au32[40/4];
1313	} Buf;
1314
1315	/*
1316	* Check the GSO properties of the socket buffer and make sure it fits.
1317	*/
1318	/** @todo Figure out how to handle SKB_GSO_TCP_ECN! */
1319	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_type & ~(SKB_GSO_DODGY \| SKB_GSO_TCPV6 \| SKB_GSO_TCPV4) ))
1320	{
1321	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_type=%#x\n", skb_shinfo(pSkb)->gso_type));
1322	return false;
1323	}
1324	if (RT_UNLIKELY( skb_shinfo(pSkb)->gso_size < 1
1325	\|\| pSkb->len > VBOX_MAX_GSO_SIZE ))
1326	{
1327	Log5(("vboxNetFltLinuxCanForwardAsGso: gso_size=%#x skb_len=%#x (max=%#x)\n", skb_shinfo(pSkb)->gso_size, pSkb->len, VBOX_MAX_GSO_SIZE));
1328	return false;
1329	}
1330
1331	/*
1332	* Switch on the ethertype.
1333	*/
1334	uEtherType = pSkb->protocol;
1335	if ( uEtherType == RT_H2N_U16_C(RTNET_ETHERTYPE_VLAN)
1336	&& pSkb->mac_len == sizeof(RTNETETHERHDR) + sizeof(uint32_t))
1337	{
1338	uint16_t const *puEtherType = skb_header_pointer(pSkb, sizeof(RTNETETHERHDR) + sizeof(uint16_t), sizeof(uint16_t), &Buf);
1339	if (puEtherType)
1340	uEtherType = *puEtherType;
1341	}
1342	switch (uEtherType)
1343	{
1344	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV4):
1345	{
1346	unsigned int cbHdr;
1347	PCRTNETIPV4 pIPv4 = (PCRTNETIPV4)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv4), &Buf);
1348	if (RT_UNLIKELY(!pIPv4))
1349	{
1350	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv4 hdr\n"));
1351	return false;
1352	}
1353
1354	cbHdr = pIPv4->ip_hl * 4;
1355	cbTransport = RT_N2H_U16(pIPv4->ip_len);
1356	if (RT_UNLIKELY( cbHdr < RTNETIPV4_MIN_LEN
1357	\|\| cbHdr > cbTransport ))
1358	{
1359	Log5(("vboxNetFltLinuxCanForwardAsGso: invalid IPv4 lengths: ip_hl=%u ip_len=%u\n", pIPv4->ip_hl, RT_N2H_U16(pIPv4->ip_len)));
1360	return false;
1361	}
1362	cbTransport -= cbHdr;
1363	offTransport = pSkb->mac_len + cbHdr;
1364	uProtocol = pIPv4->ip_p;
1365	if (uProtocol == RTNETIPV4_PROT_TCP)
1366	enmGsoType = PDMNETWORKGSOTYPE_IPV4_TCP;
1367	else if (uProtocol == RTNETIPV4_PROT_UDP)
1368	enmGsoType = PDMNETWORKGSOTYPE_IPV4_UDP;
1369	else /** @todo IPv6: 4to6 tunneling */
1370	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1371	break;
1372	}
1373
1374	case RT_H2N_U16_C(RTNET_ETHERTYPE_IPV6):
1375	{
1376	PCRTNETIPV6 pIPv6 = (PCRTNETIPV6)skb_header_pointer(pSkb, pSkb->mac_len, sizeof(Buf.IPv6), &Buf);
1377	if (RT_UNLIKELY(!pIPv6))
1378	{
1379	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access IPv6 hdr\n"));
1380	return false;
1381	}
1382
1383	cbTransport = RT_N2H_U16(pIPv6->ip6_plen);
1384	offTransport = pSkb->mac_len + sizeof(RTNETIPV6);
1385	uProtocol = pIPv6->ip6_nxt;
1386	/** @todo IPv6: Dig our way out of the other headers. */
1387	if (uProtocol == RTNETIPV4_PROT_TCP)
1388	enmGsoType = PDMNETWORKGSOTYPE_IPV6_TCP;
1389	else if (uProtocol == RTNETIPV4_PROT_UDP)
1390	enmGsoType = PDMNETWORKGSOTYPE_IPV6_UDP;
1391	else
1392	enmGsoType = PDMNETWORKGSOTYPE_INVALID;
1393	break;
1394	}
1395
1396	default:
1397	Log5(("vboxNetFltLinuxCanForwardAsGso: uEtherType=%#x\n", RT_H2N_U16(uEtherType)));
1398	return false;
1399	}
1400
1401	if (enmGsoType == PDMNETWORKGSOTYPE_INVALID)
1402	{
1403	Log5(("vboxNetFltLinuxCanForwardAsGso: Unsupported protocol %d\n", uProtocol));
1404	return false;
1405	}
1406
1407	if (RT_UNLIKELY( offTransport + cbTransport <= offTransport
1408	\|\| offTransport + cbTransport > pSkb->len
1409	\|\| cbTransport < (uProtocol == RTNETIPV4_PROT_TCP ? RTNETTCP_MIN_LEN : RTNETUDP_MIN_LEN)) )
1410	{
1411	Log5(("vboxNetFltLinuxCanForwardAsGso: Bad transport length; off=%#x + cb=%#x => %#x; skb_len=%#x (%s)\n",
1412	offTransport, cbTransport, offTransport + cbTransport, pSkb->len, PDMNetGsoTypeName(enmGsoType) ));
1413	return false;
1414	}
1415
1416	/*
1417	* Check the TCP/UDP bits.
1418	*/
1419	if (uProtocol == RTNETIPV4_PROT_TCP)
1420	{
1421	PCRTNETTCP pTcp = (PCRTNETTCP)skb_header_pointer(pSkb, offTransport, sizeof(Buf.Tcp), &Buf);
1422	if (RT_UNLIKELY(!pTcp))
1423	{
1424	Log5(("vboxNetFltLinuxCanForwardAsGso: failed to access TCP hdr\n"));
1425	return false;
1426	}
1427
1428	cbTransportHdr = pTcp->th_off * 4;
1429	pGsoCtx->cbHdrsSeg = offTransport + cbTransportHdr;
1430	if (RT_UNLIKELY( cbTransportHdr < RTNETTCP_MIN_LEN
1431	\|\| cbTransportHdr > cbTransport
1432	\|\| offTransport + cbTransportHdr >= UINT8_MAX
1433	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1434	{
1435	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for TCP header; off=%#x cb=%#x skb_len=%#x\n", offTransport, cbTransportHdr, pSkb->len));
1436	return false;
1437	}
1438
1439	}
1440	else
1441	{
1442	Assert(uProtocol == RTNETIPV4_PROT_UDP);
1443	cbTransportHdr = sizeof(RTNETUDP);
1444	pGsoCtx->cbHdrsSeg = offTransport; /* Exclude UDP header */
1445	if (RT_UNLIKELY( offTransport + cbTransportHdr >= UINT8_MAX
1446	\|\| offTransport + cbTransportHdr >= pSkb->len ))
1447	{
1448	Log5(("vboxNetFltLinuxCanForwardAsGso: No space for UDP header; off=%#x skb_len=%#x\n", offTransport, pSkb->len));
1449	return false;
1450	}
1451	}
1452
1453	/*
1454	* We're good, init the GSO context.
1455	*/
1456	pGsoCtx->u8Type = enmGsoType;
1457	pGsoCtx->cbHdrsTotal = offTransport + cbTransportHdr;
1458	pGsoCtx->cbMaxSeg = skb_shinfo(pSkb)->gso_size;
1459	pGsoCtx->offHdr1 = pSkb->mac_len;
1460	pGsoCtx->offHdr2 = offTransport;
1461	pGsoCtx->u8Unused = 0;
1462
1463	return true;
1464	}
1465
1466	/**
1467	* Forward the socket buffer as a GSO internal network frame.
1468	*
1469	* @returns IPRT status code.
1470	* @param pThis The net filter instance.
1471	* @param pSkb The GSO socket buffer.
1472	* @param fSrc The source.
1473	* @param pGsoCtx Where to return the GSO context on success.
1474	*/
1475	static int vboxNetFltLinuxForwardAsGso(PVBOXNETFLTINS pThis, struct sk_buff *pSkb, uint32_t fSrc, PCPDMNETWORKGSO pGsoCtx)
1476	{
1477	int rc;
1478	unsigned cbExtra;
1479	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pSkb, &cbExtra);
1480	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1481	if (RT_LIKELY(pSG))
1482	{
1483	vboxNetFltLinuxSkBufToSG(pThis, pSkb, pSG, cbExtra, cSegs, fSrc, pGsoCtx);
1484
1485	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1486	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1487
1488	vboxNetFltLinuxDestroySG(pSG, pSkb);
1489	rc = VINF_SUCCESS;
1490	}
1491	else
1492	{
1493	Log(("VBoxNetFlt: Dropping the sk_buff (failure case).\n"));
1494	rc = VERR_NO_MEMORY;
1495	}
1496	return rc;
1497	}
1498
1499	#endif /* VBOXNETFLT_WITH_GSO_RECV */
1500
1501	/**
1502	* Worker for vboxNetFltLinuxForwardToIntNet.
1503	*
1504	* @returns VINF_SUCCESS or VERR_NO_MEMORY.
1505	* @param pThis The net filter instance.
1506	* @param pBuf The socket buffer.
1507	* @param fSrc The source.
1508	*/
1509	static int vboxNetFltLinuxForwardSegment(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1510	{
1511	int rc;
1512	unsigned cbExtra;
1513	unsigned cSegs = vboxNetFltLinuxCalcSGSegments(pBuf, &cbExtra);
1514	PINTNETSG pSG = (PINTNETSG)alloca(RT_UOFFSETOF_DYN(INTNETSG, aSegs[cSegs]) + cbExtra);
1515	if (RT_LIKELY(pSG))
1516	{
1517	vboxNetFltLinuxSkBufToSG(pThis, pBuf, pSG, cbExtra, cSegs, fSrc, NULL /pGsoCtx/);
1518
1519	vboxNetFltDumpPacket(pSG, false, (fSrc & INTNETTRUNKDIR_HOST) ? "host" : "wire", 1);
1520	pThis->pSwitchPort->pfnRecv(pThis->pSwitchPort, NULL /* pvIf */, pSG, fSrc);
1521
1522	vboxNetFltLinuxDestroySG(pSG, pBuf);
1523	rc = VINF_SUCCESS;
1524	}
1525	else
1526	{
1527	Log(("VBoxNetFlt: Failed to allocate SG buffer.\n"));
1528	rc = VERR_NO_MEMORY;
1529	}
1530	return rc;
1531	}
1532
1533
1534	/**
1535	* I won't disclose what I do, figure it out yourself, including pThis referencing.
1536	*
1537	* @param pThis The net filter instance.
1538	* @param pBuf The socket buffer.
1539	* @param fSrc Where the packet comes from.
1540	*/
1541	static void vboxNetFltLinuxForwardToIntNetInner(PVBOXNETFLTINS pThis, struct sk_buff *pBuf, uint32_t fSrc)
1542	{
1543	#ifdef VBOXNETFLT_WITH_GSO
1544	if (skb_is_gso(pBuf))
1545	{
1546	PDMNETWORKGSO GsoCtx;
1547	Log6(("vboxNetFltLinuxForwardToIntNetInner: skb len=%u data_len=%u truesize=%u next=%p"
1548	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x ip_summed=%d\n",
1549	pBuf->len, pBuf->data_len, pBuf->truesize, pBuf->next,
1550	skb_shinfo(pBuf)->nr_frags, skb_shinfo(pBuf)->gso_size,
1551	skb_shinfo(pBuf)->gso_segs, skb_shinfo(pBuf)->gso_type,
1552	skb_shinfo(pBuf)->frag_list, pBuf->pkt_type, pBuf->ip_summed));
1553
1554	if (RT_LIKELY(fSrc & INTNETTRUNKDIR_HOST))
1555	{
1556	/*
1557	* skb_gso_segment does the following. Do we need to do it as well?
1558	*/
1559	# if RTLNX_VER_MIN(2,6,22)
1560	skb_reset_mac_header(pBuf);
1561	pBuf->mac_len = pBuf->network_header - pBuf->mac_header;
1562	# else
1563	pBuf->mac.raw = pBuf->data;
1564	pBuf->mac_len = pBuf->nh.raw - pBuf->data;
1565	# endif
1566	}
1567
1568	# ifdef VBOXNETFLT_WITH_GSO_RECV
1569	if ( (skb_shinfo(pBuf)->gso_type & (SKB_GSO_TCPV6 \| SKB_GSO_TCPV4))
1570	&& vboxNetFltLinuxCanForwardAsGso(pThis, pBuf, fSrc, &GsoCtx) )
1571	vboxNetFltLinuxForwardAsGso(pThis, pBuf, fSrc, &GsoCtx);
1572	else
1573	# endif /* VBOXNETFLT_WITH_GSO_RECV */
1574	{
1575	/* Need to segment the packet */
1576	struct sk_buff *pNext;
1577	struct sk_buff pSegment = skb_gso_segment(pBuf, 0 /supported features*/);
1578	if (IS_ERR(pSegment))
1579	{
1580	LogRel(("VBoxNetFlt: Failed to segment a packet (%d).\n", PTR_ERR(pSegment)));
1581	return;
1582	}
1583
1584	for (; pSegment; pSegment = pNext)
1585	{
1586	Log6(("vboxNetFltLinuxForwardToIntNetInner: segment len=%u data_len=%u truesize=%u next=%p"
1587	" nr_frags=%u gso_size=%u gso_seqs=%u gso_type=%x frag_list=%p pkt_type=%x\n",
1588	pSegment->len, pSegment->data_len, pSegment->truesize, pSegment->next,
1589	skb_shinfo(pSegment)->nr_frags, skb_shinfo(pSegment)->gso_size,
1590	skb_shinfo(pSegment)->gso_segs, skb_shinfo(pSegment)->gso_type,
1591	skb_shinfo(pSegment)->frag_list, pSegment->pkt_type));
1592	pNext = pSegment->next;
1593	pSegment->next = 0;
1594	vboxNetFltLinuxForwardSegment(pThis, pSegment, fSrc);
1595	dev_kfree_skb(pSegment);
1596	}
1597	}
1598	}
1599	else
1600	#endif /* VBOXNETFLT_WITH_GSO */
1601	{
1602	Log6(("vboxNetFltLinuxForwardToIntNetInner: ptk_type=%d ip_summed=%d len=%d"
1603	" data_len=%d headroom=%d hdr_len=%d csum_offset=%d\n",
1604	pBuf->pkt_type, pBuf->ip_summed, pBuf->len, pBuf->data_len, skb_headroom(pBuf),
1605	skb_headlen(pBuf), vboxNetFltLinuxGetChecksumStartOffset(pBuf)));
1606	#ifndef VBOXNETFLT_SG_SUPPORT
1607	if (pBuf->ip_summed == CHECKSUM_PARTIAL && pBuf->pkt_type == PACKET_OUTGOING)
1608	{
1609	# if RTLNX_VER_MIN(2,6,19)
1610	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1611	# else
1612	/*
1613	* Try to work around the problem with CentOS 4.7 and 5.2 (2.6.9
1614	* and 2.6.18 kernels), they pass wrong 'h' pointer down. We take IP
1615	* header length from the header itself and reconstruct 'h' pointer
1616	* to TCP (or whatever) header.
1617	*/
1618	unsigned char *tmp = pBuf->h.raw;
1619	if (pBuf->h.raw == pBuf->nh.raw && pBuf->protocol == htons(ETH_P_IP))
1620	pBuf->h.raw = pBuf->nh.raw + pBuf->nh.iph->ihl * 4;
1621	int rc = VBOX_SKB_CHECKSUM_HELP(pBuf);
1622	/* Restore the original (wrong) pointer. */
1623	pBuf->h.raw = tmp;
1624	# endif
1625	if (rc)
1626	{
1627	LogRel(("VBoxNetFlt: Failed to compute checksum, dropping the packet.\n"));
1628	return;
1629	}
1630	}
1631	#endif /* !VBOXNETFLT_SG_SUPPORT */
1632	vboxNetFltLinuxForwardSegment(pThis, pBuf, fSrc);
1633	}
1634	}
1635
1636
1637	/**
1638	* Temporarily adjust pBuf->data so it always points to the Ethernet header,
1639	* then forward it to the internal network.
1640	*
1641	* @param pThis The net filter instance.
1642	* @param pBuf The socket buffer. This is consumed by this function.
1643	*/
1644	static void vboxNetFltLinuxForwardToIntNet(PVBOXNETFLTINS pThis, struct sk_buff *pBuf)
1645	{
1646	uint32_t fSrc = pBuf->pkt_type == PACKET_OUTGOING ? INTNETTRUNKDIR_HOST : INTNETTRUNKDIR_WIRE;
1647
1648	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1649	{
1650	/*
1651	* The packet came from the wire and the driver has already consumed
1652	* mac header. We need to restore it back. Moreover, after we are
1653	* through with this skb we need to restore its original state!
1654	*/
1655	skb_push(pBuf, pBuf->mac_len);
1656	Log5(("vboxNetFltLinuxForwardToIntNet: mac_len=%d data=%p mac_header=%p network_header=%p\n",
1657	pBuf->mac_len, pBuf->data, skb_mac_header(pBuf), skb_network_header(pBuf)));
1658	}
1659
1660	vboxNetFltLinuxForwardToIntNetInner(pThis, pBuf, fSrc);
1661
1662	/*
1663	* Restore the original state of skb as there are other handlers this skb
1664	* will be provided to.
1665	*/
1666	if (RT_UNLIKELY(fSrc & INTNETTRUNKDIR_WIRE))
1667	skb_pull(pBuf, pBuf->mac_len);
1668
1669	dev_kfree_skb(pBuf);
1670	}
1671
1672
1673	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1674	/**
1675	* Work queue handler that forwards the socket buffers queued by
1676	* vboxNetFltLinuxPacketHandler to the internal network.
1677	*
1678	* @param pWork The work queue.
1679	*/
1680	# if RTLNX_VER_MIN(2,6,20)
1681	static void vboxNetFltLinuxXmitTask(struct work_struct *pWork)
1682	# else
1683	static void vboxNetFltLinuxXmitTask(void *pWork)
1684	# endif
1685	{
1686	PVBOXNETFLTINS pThis = VBOX_FLT_XT_TO_INST(pWork);
1687	struct sk_buff *pBuf;
1688
1689	Log6(("vboxNetFltLinuxXmitTask: Got work %p.\n", pWork));
1690
1691	/*
1692	* Active? Retain the instance and increment the busy counter.
1693	*/
1694	if (vboxNetFltTryRetainBusyActive(pThis))
1695	{
1696	while ((pBuf = skb_dequeue(&pThis->u.s.XmitQueue)) != NULL)
1697	vboxNetFltLinuxForwardToIntNet(pThis, pBuf);
1698
1699	vboxNetFltRelease(pThis, true /* fBusy */);
1700	}
1701	else
1702	{
1703	/** @todo Shouldn't we just drop the packets here? There is little point in
1704	* making them accumulate when the VM is paused and it'll only waste
1705	* kernel memory anyway... Hmm. maybe wait a short while (2-5 secs)
1706	* before start draining the packets (goes for the intnet ring buf
1707	* too)? */
1708	}
1709	}
1710	#endif /* !VBOXNETFLT_LINUX_NO_XMIT_QUEUE */
1711
1712	/**
1713	* Reports the GSO capabilities of the hardware NIC.
1714	*
1715	* @param pThis The net filter instance. The caller hold a
1716	* reference to this.
1717	*/
1718	static void vboxNetFltLinuxReportNicGsoCapabilities(PVBOXNETFLTINS pThis)
1719	{
1720	#if defined(VBOXNETFLT_WITH_GSO_XMIT_WIRE) \|\| defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
1721	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1722	{
1723	struct net_device *pDev;
1724	unsigned int fFeatures;
1725
1726	RTSpinlockAcquire(pThis->hSpinlock);
1727
1728	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1729	if (pDev)
1730	fFeatures = pDev->features;
1731	else
1732	fFeatures = 0;
1733
1734	RTSpinlockRelease(pThis->hSpinlock);
1735
1736	if (pThis->pSwitchPort)
1737	{
1738	/* Set/update the GSO capabilities of the NIC. */
1739	uint32_t fGsoCapabilites = 0;
1740	if (fFeatures & NETIF_F_TSO)
1741	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP);
1742	if (fFeatures & NETIF_F_TSO6)
1743	fGsoCapabilites \|= RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP);
1744	Log3(("vboxNetFltLinuxReportNicGsoCapabilities: reporting wire %s%s\n",
1745	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)) ? "tso " : "",
1746	(fGsoCapabilites & RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)) ? "tso6 " : ""));
1747	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort, fGsoCapabilites, INTNETTRUNKDIR_WIRE);
1748	}
1749
1750	vboxNetFltRelease(pThis, true /fBusy/);
1751	}
1752	#endif /* VBOXNETFLT_WITH_GSO_XMIT_WIRE \|\| VBOXNETFLT_WITH_GSO_XMIT_HOST */
1753	}
1754
1755	/**
1756	* Helper that determines whether the host (ignoreing us) is operating the
1757	* interface in promiscuous mode or not.
1758	*/
1759	static bool vboxNetFltLinuxPromiscuous(PVBOXNETFLTINS pThis)
1760	{
1761	bool fRc = false;
1762	struct net_device * pDev = vboxNetFltLinuxRetainNetDev(pThis);
1763	if (pDev)
1764	{
1765	fRc = !!(pDev->promiscuity - (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet) & 1));
1766	LogFlow(("vboxNetFltPortOsIsPromiscuous: returns %d, pDev->promiscuity=%d, fPromiscuousSet=%d\n",
1767	fRc, pDev->promiscuity, pThis->u.s.fPromiscuousSet));
1768	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
1769	}
1770	return fRc;
1771	}
1772
1773	/**
1774	* Does this device needs link state change signaled?
1775	* Currently we need it for our own VBoxNetAdp and TAP.
1776	*/
1777	static bool vboxNetFltNeedsLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev)
1778	{
1779	if (pDev->ethtool_ops && pDev->ethtool_ops->get_drvinfo)
1780	{
1781	struct ethtool_drvinfo Info;
1782
1783	memset(&Info, 0, sizeof(Info));
1784	Info.cmd = ETHTOOL_GDRVINFO;
1785	pDev->ethtool_ops->get_drvinfo(pDev, &Info);
1786	Log3(("%s: driver=%.s version=%.s bus_info=%.*s\n",
1787	__FUNCTION__,
1788	sizeof(Info.driver), Info.driver,
1789	sizeof(Info.version), Info.version,
1790	sizeof(Info.bus_info), Info.bus_info));
1791
1792	if (!strncmp(Info.driver, "vboxnet", sizeof(Info.driver)))
1793	return true;
1794
1795	#if RTLNX_VER_MIN(2,6,36) /* TAP started doing carrier */
1796	return !strncmp(Info.driver, "tun", 4)
1797	&& !strncmp(Info.bus_info, "tap", 4);
1798	#endif
1799	}
1800
1801	return false;
1802	}
1803
1804	#if RTLNX_VER_MAX(2,6,18)
1805	DECLINLINE(void) netif_tx_lock_bh(struct net_device *pDev)
1806	{
1807	spin_lock_bh(&pDev->xmit_lock);
1808	}
1809
1810	DECLINLINE(void) netif_tx_unlock_bh(struct net_device *pDev)
1811	{
1812	spin_unlock_bh(&pDev->xmit_lock);
1813	}
1814	#endif
1815
1816	/**
1817	* Some devices need link state change when filter attaches/detaches
1818	* since the filter is their link in a sense.
1819	*/
1820	static void vboxNetFltSetLinkState(PVBOXNETFLTINS pThis, struct net_device *pDev, bool fLinkUp)
1821	{
1822	if (vboxNetFltNeedsLinkState(pThis, pDev))
1823	{
1824	Log3(("%s: bringing device link %s\n",
1825	__FUNCTION__, fLinkUp ? "up" : "down"));
1826	netif_tx_lock_bh(pDev);
1827	if (fLinkUp)
1828	netif_carrier_on(pDev);
1829	else
1830	netif_carrier_off(pDev);
1831	netif_tx_unlock_bh(pDev);
1832	}
1833	}
1834
1835	/**
1836	* Internal worker for vboxNetFltLinuxNotifierCallback.
1837	*
1838	* @returns VBox status code.
1839	* @param pThis The instance.
1840	* @param pDev The device to attach to.
1841	*/
1842	static int vboxNetFltLinuxAttachToInterface(PVBOXNETFLTINS pThis, struct net_device *pDev)
1843	{
1844	LogFlow(("vboxNetFltLinuxAttachToInterface: pThis=%p (%s)\n", pThis, pThis->szName));
1845
1846	/*
1847	* Retain and store the device.
1848	*/
1849	dev_hold(pDev);
1850
1851	RTSpinlockAcquire(pThis->hSpinlock);
1852	ASMAtomicUoWritePtr(&pThis->u.s.pDev, pDev);
1853	RTSpinlockRelease(pThis->hSpinlock);
1854
1855	Log(("vboxNetFltLinuxAttachToInterface: Device %p(%s) retained. ref=%d\n",
1856	pDev, pDev->name,
1857	#if RTLNX_VER_MIN(2,6,37)
1858	netdev_refcnt_read(pDev)
1859	#else
1860	atomic_read(&pDev->refcnt)
1861	#endif
1862	));
1863	Log(("vboxNetFltLinuxAttachToInterface: Got pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
1864	pDev, pThis, ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *)));
1865
1866	/* Get the mac address while we still have a valid net_device reference. */
1867	memcpy(&pThis->u.s.MacAddr, pDev->dev_addr, sizeof(pThis->u.s.MacAddr));
1868	/* Initialize MTU */
1869	pThis->u.s.cbMtu = pDev->mtu;
1870
1871	/*
1872	* Install a packet filter for this device with a protocol wildcard (ETH_P_ALL).
1873	*/
1874	pThis->u.s.PacketType.type = __constant_htons(ETH_P_ALL);
1875	pThis->u.s.PacketType.dev = pDev;
1876	pThis->u.s.PacketType.func = vboxNetFltLinuxPacketHandler;
1877	dev_add_pack(&pThis->u.s.PacketType);
1878	ASMAtomicUoWriteBool(&pThis->u.s.fPacketHandler, true);
1879	Log(("vboxNetFltLinuxAttachToInterface: this=%p: Packet handler installed.\n", pThis));
1880
1881	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1882	vboxNetFltLinuxHookDev(pThis, pDev);
1883	#endif
1884
1885	/*
1886	* Are we the "carrier" for this device (e.g. vboxnet or tap)?
1887	*/
1888	vboxNetFltSetLinkState(pThis, pDev, true);
1889
1890	/*
1891	* Set indicators that require the spinlock. Be abit paranoid about racing
1892	* the device notification handle.
1893	*/
1894	RTSpinlockAcquire(pThis->hSpinlock);
1895	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
1896	if (pDev)
1897	{
1898	ASMAtomicUoWriteBool(&pThis->fDisconnectedFromHost, false);
1899	ASMAtomicUoWriteBool(&pThis->u.s.fRegistered, true);
1900	pDev = NULL; /* don't dereference it */
1901	}
1902	RTSpinlockRelease(pThis->hSpinlock);
1903
1904	/*
1905	* Report GSO capabilities
1906	*/
1907	Assert(pThis->pSwitchPort);
1908	if (vboxNetFltTryRetainBusyNotDisconnected(pThis))
1909	{
1910	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
1911	pThis->pSwitchPort->pfnReportMacAddress(pThis->pSwitchPort, &pThis->u.s.MacAddr);
1912	pThis->pSwitchPort->pfnReportPromiscuousMode(pThis->pSwitchPort, vboxNetFltLinuxPromiscuous(pThis));
1913	pThis->pSwitchPort->pfnReportNoPreemptDsts(pThis->pSwitchPort, INTNETTRUNKDIR_WIRE \| INTNETTRUNKDIR_HOST);
1914	vboxNetFltRelease(pThis, true /fBusy/);
1915	}
1916
1917	LogRel(("VBoxNetFlt: attached to '%s' / %RTmac\n", pThis->szName, &pThis->u.s.MacAddr));
1918	return VINF_SUCCESS;
1919	}
1920
1921
1922	static int vboxNetFltLinuxUnregisterDevice(PVBOXNETFLTINS pThis, struct net_device *pDev)
1923	{
1924	bool fRegistered;
1925	Assert(!pThis->fDisconnectedFromHost);
1926
1927	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
1928	vboxNetFltLinuxUnhookDev(pThis, pDev);
1929	#endif
1930
1931	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
1932	{
1933	dev_remove_pack(&pThis->u.s.PacketType);
1934	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: packet handler removed.\n", pThis));
1935	}
1936
1937	RTSpinlockAcquire(pThis->hSpinlock);
1938	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
1939	if (fRegistered)
1940	{
1941	ASMAtomicWriteBool(&pThis->fDisconnectedFromHost, true);
1942	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
1943	}
1944	RTSpinlockRelease(pThis->hSpinlock);
1945
1946	if (fRegistered)
1947	{
1948	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
1949	skb_queue_purge(&pThis->u.s.XmitQueue);
1950	#endif
1951	Log(("vboxNetFltLinuxUnregisterDevice: this=%p: xmit queue purged.\n", pThis));
1952	Log(("vboxNetFltLinuxUnregisterDevice: Device %p(%s) released. ref=%d\n",
1953	pDev, pDev->name,
1954	#if RTLNX_VER_MIN(2,6,37)
1955	netdev_refcnt_read(pDev)
1956	#else
1957	atomic_read(&pDev->refcnt)
1958	#endif
1959	));
1960	dev_put(pDev);
1961	}
1962
1963	return NOTIFY_OK;
1964	}
1965
1966	static int vboxNetFltLinuxDeviceIsUp(PVBOXNETFLTINS pThis, struct net_device *pDev)
1967	{
1968	/* Check if we are not suspended and promiscuous mode has not been set. */
1969	if ( pThis->enmTrunkState == INTNETTRUNKIFSTATE_ACTIVE
1970	&& !ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1971	{
1972	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1973	dev_set_promiscuity(pDev, 1);
1974	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, true);
1975	Log(("vboxNetFltLinuxDeviceIsUp: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1976	}
1977	else
1978	Log(("vboxNetFltLinuxDeviceIsUp: no need to enable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1979	return NOTIFY_OK;
1980	}
1981
1982	static int vboxNetFltLinuxDeviceGoingDown(PVBOXNETFLTINS pThis, struct net_device *pDev)
1983	{
1984	/* Undo promiscuous mode if we has set it. */
1985	if (ASMAtomicUoReadBool(&pThis->u.s.fPromiscuousSet))
1986	{
1987	/* Note that there is no need for locking as the kernel got hold of the lock already. */
1988	dev_set_promiscuity(pDev, -1);
1989	ASMAtomicWriteBool(&pThis->u.s.fPromiscuousSet, false);
1990	Log(("vboxNetFltLinuxDeviceGoingDown: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1991	}
1992	else
1993	Log(("vboxNetFltLinuxDeviceGoingDown: no need to disable promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
1994	return NOTIFY_OK;
1995	}
1996
1997	/**
1998	* Callback for listening to MTU change event.
1999	*
2000	* We need to track changes of host's inteface MTU to discard over-sized frames
2001	* coming from the internal network as they may hang the TX queue of host's
2002	* adapter.
2003	*
2004	* @returns NOTIFY_OK
2005	* @param pThis The netfilter instance.
2006	* @param pDev Pointer to device structure of host's interface.
2007	*/
2008	static int vboxNetFltLinuxDeviceMtuChange(PVBOXNETFLTINS pThis, struct net_device *pDev)
2009	{
2010	ASMAtomicWriteU32(&pThis->u.s.cbMtu, pDev->mtu);
2011	Log(("vboxNetFltLinuxDeviceMtuChange: set MTU for %s to %d\n", pThis->szName, pDev->mtu));
2012	return NOTIFY_OK;
2013	}
2014
2015	#ifdef LOG_ENABLED
2016	/** Stringify the NETDEV_XXX constants. */
2017	static const char *vboxNetFltLinuxGetNetDevEventName(unsigned long ulEventType)
2018	{
2019	const char *pszEvent = "NETDEV_<unknown>";
2020	switch (ulEventType)
2021	{
2022	case NETDEV_REGISTER: pszEvent = "NETDEV_REGISTER"; break;
2023	case NETDEV_UNREGISTER: pszEvent = "NETDEV_UNREGISTER"; break;
2024	case NETDEV_UP: pszEvent = "NETDEV_UP"; break;
2025	case NETDEV_DOWN: pszEvent = "NETDEV_DOWN"; break;
2026	case NETDEV_REBOOT: pszEvent = "NETDEV_REBOOT"; break;
2027	case NETDEV_CHANGENAME: pszEvent = "NETDEV_CHANGENAME"; break;
2028	case NETDEV_CHANGE: pszEvent = "NETDEV_CHANGE"; break;
2029	case NETDEV_CHANGEMTU: pszEvent = "NETDEV_CHANGEMTU"; break;
2030	case NETDEV_CHANGEADDR: pszEvent = "NETDEV_CHANGEADDR"; break;
2031	case NETDEV_GOING_DOWN: pszEvent = "NETDEV_GOING_DOWN"; break;
2032	# ifdef NETDEV_FEAT_CHANGE
2033	case NETDEV_FEAT_CHANGE: pszEvent = "NETDEV_FEAT_CHANGE"; break;
2034	# endif
2035	}
2036	return pszEvent;
2037	}
2038	#endif /* LOG_ENABLED */
2039
2040	/**
2041	* Callback for listening to netdevice events.
2042	*
2043	* This works the rediscovery, clean up on unregistration, promiscuity on
2044	* up/down, and GSO feature changes from ethtool.
2045	*
2046	* @returns NOTIFY_OK
2047	* @param self Pointer to our notifier registration block.
2048	* @param ulEventType The event.
2049	* @param ptr Event specific, but it is usually the device it
2050	* relates to.
2051	*/
2052	static int vboxNetFltLinuxNotifierCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2053
2054	{
2055	PVBOXNETFLTINS pThis = VBOX_FLT_NB_TO_INST(self);
2056	struct net_device pMyDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device );
2057	struct net_device *pDev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2058	int rc = NOTIFY_OK;
2059
2060	Log(("VBoxNetFlt: got event %s(0x%lx) on %s, pDev=%p pThis=%p pThis->u.s.pDev=%p\n",
2061	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType, pDev->name, pDev, pThis, pMyDev));
2062
2063	if (ulEventType == NETDEV_REGISTER)
2064	{
2065	#if RTLNX_VER_MIN(2,6,24) /* cgroups/namespaces introduced */
2066	# if RTLNX_VER_MIN(2,6,26)
2067	# define VBOX_DEV_NET(dev) dev_net(dev)
2068	# define VBOX_NET_EQ(n1, n2) net_eq((n1), (n2))
2069	# else
2070	# define VBOX_DEV_NET(dev) ((dev)->nd_net)
2071	# define VBOX_NET_EQ(n1, n2) ((n1) == (n2))
2072	# endif
2073	struct net *pMyNet = current->nsproxy->net_ns;
2074	struct net *pDevNet = VBOX_DEV_NET(pDev);
2075
2076	if (VBOX_NET_EQ(pDevNet, pMyNet))
2077	#endif /* namespaces */
2078	{
2079	if (strcmp(pDev->name, pThis->szName) == 0)
2080	{
2081	vboxNetFltLinuxAttachToInterface(pThis, pDev);
2082	}
2083	}
2084	}
2085	else
2086	{
2087	if (pDev == pMyDev)
2088	{
2089	switch (ulEventType)
2090	{
2091	case NETDEV_UNREGISTER:
2092	rc = vboxNetFltLinuxUnregisterDevice(pThis, pDev);
2093	break;
2094	case NETDEV_UP:
2095	rc = vboxNetFltLinuxDeviceIsUp(pThis, pDev);
2096	break;
2097	case NETDEV_GOING_DOWN:
2098	rc = vboxNetFltLinuxDeviceGoingDown(pThis, pDev);
2099	break;
2100	case NETDEV_CHANGEMTU:
2101	rc = vboxNetFltLinuxDeviceMtuChange(pThis, pDev);
2102	break;
2103	case NETDEV_CHANGENAME:
2104	break;
2105	#ifdef NETDEV_FEAT_CHANGE
2106	case NETDEV_FEAT_CHANGE:
2107	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2108	break;
2109	#endif
2110	}
2111	}
2112	}
2113
2114	return rc;
2115	}
2116
2117	/*
2118	* Initial enumeration of netdevs. Called with NETDEV_REGISTER by
2119	* register_netdevice_notifier() under rtnl lock.
2120	*/
2121	static int vboxNetFltLinuxEnumeratorCallback(struct notifier_block self, unsigned long ulEventType, void ptr)
2122	{
2123	PVBOXNETFLTINS pThis = ((PVBOXNETFLTNOTIFIER)self)->pThis;
2124	struct net_device *dev = VBOX_NETDEV_NOTIFIER_INFO_TO_DEV(ptr);
2125	struct in_device *in_dev;
2126	struct inet6_dev *in6_dev;
2127
2128	if (ulEventType != NETDEV_REGISTER)
2129	return NOTIFY_OK;
2130
2131	if (RT_UNLIKELY(pThis->pSwitchPort->pfnNotifyHostAddress == NULL))
2132	return NOTIFY_OK;
2133
2134	/*
2135	* IPv4
2136	*/
2137	#if RTLNX_VER_MIN(2,6,14)
2138	in_dev = __in_dev_get_rtnl(dev);
2139	#else
2140	in_dev = __in_dev_get(dev);
2141	#endif
2142	if (in_dev != NULL)
2143	{
2144	struct in_ifaddr *ifa;
2145
2146	for (ifa = in_dev->ifa_list; ifa; ifa = ifa->ifa_next) {
2147	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2148	return NOTIFY_OK;
2149
2150	if ( dev != pThis->u.s.pDev
2151	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2152	continue;
2153
2154	Log(("%s: %s: IPv4 addr %RTnaipv4 mask %RTnaipv4\n",
2155	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2156	ifa->ifa_address, ifa->ifa_mask));
2157
2158	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2159	/* :fAdded */ true, kIntNetAddrType_IPv4, &ifa->ifa_address);
2160	}
2161	}
2162
2163	/*
2164	* IPv6
2165	*/
2166	in6_dev = __in6_dev_get(dev);
2167	if (in6_dev != NULL)
2168	{
2169	struct inet6_ifaddr *ifa;
2170
2171	read_lock_bh(&in6_dev->lock);
2172	#if RTLNX_VER_MIN(2,6,35)
2173	list_for_each_entry(ifa, &in6_dev->addr_list, if_list)
2174	#else
2175	for (ifa = in6_dev->addr_list; ifa != NULL; ifa = ifa->if_next)
2176	#endif
2177	{
2178	if ( dev != pThis->u.s.pDev
2179	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2180	continue;
2181
2182	Log(("%s: %s: IPv6 addr %RTnaipv6/%u\n",
2183	__FUNCTION__, VBOX_NETDEV_NAME(dev),
2184	&ifa->addr, (unsigned)ifa->prefix_len));
2185
2186	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort,
2187	/* :fAdded */ true, kIntNetAddrType_IPv6, &ifa->addr);
2188	}
2189	read_unlock_bh(&in6_dev->lock);
2190	}
2191
2192	return NOTIFY_OK;
2193	}
2194
2195
2196	static int vboxNetFltLinuxNotifierIPv4Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2197	{
2198	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv4);
2199	struct net_device pDev, pEventDev;
2200	struct in_ifaddr ifa = (struct in_ifaddr )ptr;
2201	bool fMyDev;
2202	int rc = NOTIFY_OK;
2203
2204	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2205	pEventDev = ifa->ifa_dev->dev;
2206	fMyDev = (pDev == pEventDev);
2207	Log(("VBoxNetFlt: %s: IPv4 event %s(0x%lx) %s: addr %RTnaipv4 mask %RTnaipv4\n",
2208	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2209	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2210	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2211	ifa->ifa_address, ifa->ifa_mask));
2212
2213	if (pDev != NULL)
2214	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2215
2216	if (VBOX_IPV4_IS_LOOPBACK(ifa->ifa_address))
2217	return NOTIFY_OK;
2218
2219	if ( !fMyDev
2220	&& VBOX_IPV4_IS_LINKLOCAL_169(ifa->ifa_address))
2221	return NOTIFY_OK;
2222
2223	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2224	{
2225	bool fAdded;
2226	if (ulEventType == NETDEV_UP)
2227	fAdded = true;
2228	else if (ulEventType == NETDEV_DOWN)
2229	fAdded = false;
2230	else
2231	return NOTIFY_OK;
2232
2233	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2234	kIntNetAddrType_IPv4, &ifa->ifa_local);
2235	}
2236
2237	return rc;
2238	}
2239
2240
2241	static int vboxNetFltLinuxNotifierIPv6Callback(struct notifier_block self, unsigned long ulEventType, void ptr)
2242	{
2243	PVBOXNETFLTINS pThis = RT_FROM_MEMBER(self, VBOXNETFLTINS, u.s.NotifierIPv6);
2244	struct net_device pDev, pEventDev;
2245	struct inet6_ifaddr ifa = (struct inet6_ifaddr )ptr;
2246	bool fMyDev;
2247	int rc = NOTIFY_OK;
2248
2249	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2250	pEventDev = ifa->idev->dev;
2251	fMyDev = (pDev == pEventDev);
2252	Log(("VBoxNetFlt: %s: IPv6 event %s(0x%lx) %s: %RTnaipv6\n",
2253	pDev ? VBOX_NETDEV_NAME(pDev) : "<unknown>",
2254	vboxNetFltLinuxGetNetDevEventName(ulEventType), ulEventType,
2255	pEventDev ? VBOX_NETDEV_NAME(pEventDev) : "<unknown>",
2256	&ifa->addr));
2257
2258	if (pDev != NULL)
2259	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2260
2261	if ( !fMyDev
2262	&& ipv6_addr_type(&ifa->addr) & (IPV6_ADDR_LINKLOCAL \| IPV6_ADDR_LOOPBACK))
2263	return NOTIFY_OK;
2264
2265	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2266	{
2267	bool fAdded;
2268	if (ulEventType == NETDEV_UP)
2269	fAdded = true;
2270	else if (ulEventType == NETDEV_DOWN)
2271	fAdded = false;
2272	else
2273	return NOTIFY_OK;
2274
2275	pThis->pSwitchPort->pfnNotifyHostAddress(pThis->pSwitchPort, fAdded,
2276	kIntNetAddrType_IPv6, &ifa->addr);
2277	}
2278
2279	return rc;
2280	}
2281
2282
2283	bool vboxNetFltOsMaybeRediscovered(PVBOXNETFLTINS pThis)
2284	{
2285	return !ASMAtomicUoReadBool(&pThis->fDisconnectedFromHost);
2286	}
2287
2288	int vboxNetFltPortOsXmit(PVBOXNETFLTINS pThis, void *pvIfData, PINTNETSG pSG, uint32_t fDst)
2289	{
2290	struct net_device * pDev;
2291	int err;
2292	int rc = VINF_SUCCESS;
2293	IPRT_LINUX_SAVE_EFL_AC();
2294	NOREF(pvIfData);
2295
2296	LogFlow(("vboxNetFltPortOsXmit: pThis=%p (%s)\n", pThis, pThis->szName));
2297
2298	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2299	if (pDev)
2300	{
2301	/*
2302	* Create a sk_buff for the gather list and push it onto the wire.
2303	*/
2304	if (fDst & INTNETTRUNKDIR_WIRE)
2305	{
2306	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, true);
2307	if (pBuf)
2308	{
2309	vboxNetFltDumpPacket(pSG, true, "wire", 1);
2310	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2311	Log6(("vboxNetFltPortOsXmit: dev_queue_xmit(%p)\n", pBuf));
2312	err = dev_queue_xmit(pBuf);
2313	if (err)
2314	rc = RTErrConvertFromErrno(err);
2315	}
2316	else
2317	rc = VERR_NO_MEMORY;
2318	}
2319
2320	/*
2321	* Create a sk_buff for the gather list and push it onto the host stack.
2322	*/
2323	if (fDst & INTNETTRUNKDIR_HOST)
2324	{
2325	struct sk_buff *pBuf = vboxNetFltLinuxSkBufFromSG(pThis, pSG, false);
2326	if (pBuf)
2327	{
2328	vboxNetFltDumpPacket(pSG, true, "host", (fDst & INTNETTRUNKDIR_WIRE) ? 0 : 1);
2329	Log6(("vboxNetFltPortOsXmit: pBuf->cb dump:\n%.*Rhxd\n", sizeof(pBuf->cb), pBuf->cb));
2330	Log6(("vboxNetFltPortOsXmit: netif_rx_ni(%p)\n", pBuf));
2331	#if RTLNX_VER_MIN(5,18,0) \|\| RTLNX_RHEL_MIN(9,1)
2332	local_bh_disable();
2333	err = netif_rx(pBuf);
2334	local_bh_enable();
2335	#else
2336	err = netif_rx_ni(pBuf);
2337	#endif
2338	if (err)
2339	rc = RTErrConvertFromErrno(err);
2340	}
2341	else
2342	rc = VERR_NO_MEMORY;
2343	}
2344
2345	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2346	}
2347
2348	IPRT_LINUX_RESTORE_EFL_AC();
2349	return rc;
2350	}
2351
2352
2353	void vboxNetFltPortOsSetActive(PVBOXNETFLTINS pThis, bool fActive)
2354	{
2355	struct net_device *pDev;
2356	IPRT_LINUX_SAVE_EFL_AC();
2357
2358	LogFlow(("vboxNetFltPortOsSetActive: pThis=%p (%s), fActive=%RTbool, fDisablePromiscuous=%RTbool\n",
2359	pThis, pThis->szName, fActive, pThis->fDisablePromiscuous));
2360
2361	if (pThis->fDisablePromiscuous)
2362	return;
2363
2364	pDev = vboxNetFltLinuxRetainNetDev(pThis);
2365	if (pDev)
2366	{
2367	/*
2368	* This api is a bit weird, the best reference is the code.
2369	*
2370	* Also, we have a bit or race conditions wrt the maintenance of
2371	* host the interface promiscuity for vboxNetFltPortOsIsPromiscuous.
2372	*/
2373	#ifdef LOG_ENABLED
2374	u_int16_t fIf;
2375	unsigned const cPromiscBefore = pDev->promiscuity;
2376	#endif
2377	if (fActive)
2378	{
2379	Assert(!pThis->u.s.fPromiscuousSet);
2380
2381	rtnl_lock();
2382	dev_set_promiscuity(pDev, 1);
2383	rtnl_unlock();
2384	pThis->u.s.fPromiscuousSet = true;
2385	Log(("vboxNetFltPortOsSetActive: enabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2386	}
2387	else
2388	{
2389	if (pThis->u.s.fPromiscuousSet)
2390	{
2391	rtnl_lock();
2392	dev_set_promiscuity(pDev, -1);
2393	rtnl_unlock();
2394	Log(("vboxNetFltPortOsSetActive: disabled promiscuous mode on %s (%d)\n", pThis->szName, pDev->promiscuity));
2395	}
2396	pThis->u.s.fPromiscuousSet = false;
2397
2398	#ifdef LOG_ENABLED
2399	fIf = dev_get_flags(pDev);
2400	Log(("VBoxNetFlt: fIf=%#x; %d->%d\n", fIf, cPromiscBefore, pDev->promiscuity));
2401	#endif
2402	}
2403
2404	vboxNetFltLinuxReleaseNetDev(pThis, pDev);
2405	}
2406	IPRT_LINUX_RESTORE_EFL_AC();
2407	}
2408
2409
2410	int vboxNetFltOsDisconnectIt(PVBOXNETFLTINS pThis)
2411	{
2412	/*
2413	* Remove packet handler when we get disconnected from internal switch as
2414	* we don't want the handler to forward packets to disconnected switch.
2415	*/
2416	if (ASMAtomicCmpXchgBool(&pThis->u.s.fPacketHandler, false, true))
2417	{
2418	IPRT_LINUX_SAVE_EFL_AC();
2419	dev_remove_pack(&pThis->u.s.PacketType);
2420	Log(("vboxNetFltOsDisconnectIt: this=%p: Packet handler removed.\n", pThis));
2421	IPRT_LINUX_RESTORE_EFL_AC();
2422	}
2423	return VINF_SUCCESS;
2424	}
2425
2426
2427	int vboxNetFltOsConnectIt(PVBOXNETFLTINS pThis)
2428	{
2429	IPRT_LINUX_SAVE_EFL_AC();
2430
2431	/*
2432	* Report the GSO capabilities of the host and device (if connected).
2433	* Note! No need to mark ourselves busy here.
2434	*/
2435	/** @todo duplicate work here now? Attach */
2436	#if defined(VBOXNETFLT_WITH_GSO_XMIT_HOST)
2437	Log3(("vboxNetFltOsConnectIt: reporting host tso tso6\n"));
2438	pThis->pSwitchPort->pfnReportGsoCapabilities(pThis->pSwitchPort,
2439	0
2440	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV4_TCP)
2441	\| RT_BIT_32(PDMNETWORKGSOTYPE_IPV6_TCP)
2442	, INTNETTRUNKDIR_HOST);
2443
2444	#endif
2445	vboxNetFltLinuxReportNicGsoCapabilities(pThis);
2446
2447	IPRT_LINUX_RESTORE_EFL_AC();
2448	return VINF_SUCCESS;
2449	}
2450
2451
2452	void vboxNetFltOsDeleteInstance(PVBOXNETFLTINS pThis)
2453	{
2454	struct net_device *pDev;
2455	bool fRegistered;
2456	IPRT_LINUX_SAVE_EFL_AC();
2457
2458	#ifdef VBOXNETFLT_WITH_HOST2WIRE_FILTER
2459	vboxNetFltLinuxUnhookDev(pThis, NULL);
2460	#endif
2461
2462	/** @todo This code may race vboxNetFltLinuxUnregisterDevice (very very
2463	* unlikely, but none the less). Since it doesn't actually update the
2464	* state (just reads it), it is likely to panic in some interesting
2465	* ways. */
2466
2467	RTSpinlockAcquire(pThis->hSpinlock);
2468	pDev = ASMAtomicUoReadPtrT(&pThis->u.s.pDev, struct net_device *);
2469	fRegistered = ASMAtomicXchgBool(&pThis->u.s.fRegistered, false);
2470	RTSpinlockRelease(pThis->hSpinlock);
2471
2472	if (fRegistered)
2473	{
2474	vboxNetFltSetLinkState(pThis, pDev, false);
2475
2476	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2477	skb_queue_purge(&pThis->u.s.XmitQueue);
2478	#endif
2479	Log(("vboxNetFltOsDeleteInstance: this=%p: xmit queue purged.\n", pThis));
2480	Log(("vboxNetFltOsDeleteInstance: Device %p(%s) released. ref=%d\n",
2481	pDev, pDev->name,
2482	#if RTLNX_VER_MIN(2,6,37)
2483	netdev_refcnt_read(pDev)
2484	#else
2485	atomic_read(&pDev->refcnt)
2486	#endif
2487	));
2488	dev_put(pDev);
2489	}
2490
2491	unregister_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2492	unregister_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2493
2494	Log(("vboxNetFltOsDeleteInstance: this=%p: Notifier removed.\n", pThis));
2495	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2496	module_put(THIS_MODULE);
2497
2498	IPRT_LINUX_RESTORE_EFL_AC();
2499	}
2500
2501
2502	int vboxNetFltOsInitInstance(PVBOXNETFLTINS pThis, void *pvContext)
2503	{
2504	int err;
2505	IPRT_LINUX_SAVE_EFL_AC();
2506	NOREF(pvContext);
2507
2508	pThis->u.s.Notifier.notifier_call = vboxNetFltLinuxNotifierCallback;
2509	err = register_netdevice_notifier(&pThis->u.s.Notifier);
2510	if (err)
2511	{
2512	IPRT_LINUX_RESTORE_EFL_AC();
2513	return VERR_INTNET_FLT_IF_FAILED;
2514	}
2515	if (!pThis->u.s.fRegistered)
2516	{
2517	unregister_netdevice_notifier(&pThis->u.s.Notifier);
2518	LogRel(("VBoxNetFlt: failed to find %s.\n", pThis->szName));
2519	IPRT_LINUX_RESTORE_EFL_AC();
2520	return VERR_INTNET_FLT_IF_NOT_FOUND;
2521	}
2522
2523	Log(("vboxNetFltOsInitInstance: this=%p: Notifier installed.\n", pThis));
2524	if ( pThis->fDisconnectedFromHost
2525	\|\| !try_module_get(THIS_MODULE))
2526	{
2527	IPRT_LINUX_RESTORE_EFL_AC();
2528	return VERR_INTNET_FLT_IF_FAILED;
2529	}
2530
2531	if (pThis->pSwitchPort->pfnNotifyHostAddress)
2532	{
2533	VBOXNETFLTNOTIFIER Enumerator;
2534
2535	/*
2536	* register_inetaddr_notifier() and register_inet6addr_notifier()
2537	* do not call the callback for existing devices. Enumerating
2538	* all network devices explicitly is a bit of an ifdef mess,
2539	* so co-opt register_netdevice_notifier() to do that for us.
2540	*/
2541	RT_ZERO(Enumerator);
2542	Enumerator.Notifier.notifier_call = vboxNetFltLinuxEnumeratorCallback;
2543	Enumerator.pThis = pThis;
2544
2545	err = register_netdevice_notifier(&Enumerator.Notifier);
2546	if (err)
2547	{
2548	LogRel(("%s: failed to enumerate network devices: error %d\n", __FUNCTION__, err));
2549	IPRT_LINUX_RESTORE_EFL_AC();
2550	return VINF_SUCCESS;
2551	}
2552
2553	unregister_netdevice_notifier(&Enumerator.Notifier);
2554
2555	pThis->u.s.NotifierIPv4.notifier_call = vboxNetFltLinuxNotifierIPv4Callback;
2556	err = register_inetaddr_notifier(&pThis->u.s.NotifierIPv4);
2557	if (err)
2558	LogRel(("%s: failed to register IPv4 notifier: error %d\n", __FUNCTION__, err));
2559
2560	pThis->u.s.NotifierIPv6.notifier_call = vboxNetFltLinuxNotifierIPv6Callback;
2561	err = register_inet6addr_notifier(&pThis->u.s.NotifierIPv6);
2562	if (err)
2563	LogRel(("%s: failed to register IPv6 notifier: error %d\n", __FUNCTION__, err));
2564	}
2565
2566	IPRT_LINUX_RESTORE_EFL_AC();
2567	return VINF_SUCCESS;
2568	}
2569
2570	int vboxNetFltOsPreInitInstance(PVBOXNETFLTINS pThis)
2571	{
2572	IPRT_LINUX_SAVE_EFL_AC();
2573
2574	/*
2575	* Init the linux specific members.
2576	*/
2577	ASMAtomicUoWriteNullPtr(&pThis->u.s.pDev);
2578	pThis->u.s.fRegistered = false;
2579	pThis->u.s.fPromiscuousSet = false;
2580	pThis->u.s.fPacketHandler = false;
2581	memset(&pThis->u.s.PacketType, 0, sizeof(pThis->u.s.PacketType));
2582	#ifndef VBOXNETFLT_LINUX_NO_XMIT_QUEUE
2583	skb_queue_head_init(&pThis->u.s.XmitQueue);
2584	# if RTLNX_VER_MIN(2,6,20)
2585	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask);
2586	# else
2587	INIT_WORK(&pThis->u.s.XmitTask, vboxNetFltLinuxXmitTask, &pThis->u.s.XmitTask);
2588	# endif
2589	#endif
2590
2591	IPRT_LINUX_RESTORE_EFL_AC();
2592	return VINF_SUCCESS;
2593	}
2594
2595
2596	void vboxNetFltPortOsNotifyMacAddress(PVBOXNETFLTINS pThis, void *pvIfData, PCRTMAC pMac)
2597	{
2598	NOREF(pThis); NOREF(pvIfData); NOREF(pMac);
2599	}
2600
2601
2602	int vboxNetFltPortOsConnectInterface(PVBOXNETFLTINS pThis, void pvIf, void *pvIfData)
2603	{
2604	/* Nothing to do */
2605	NOREF(pThis); NOREF(pvIf); NOREF(pvIfData);
2606	return VINF_SUCCESS;
2607	}
2608
2609
2610	int vboxNetFltPortOsDisconnectInterface(PVBOXNETFLTINS pThis, void *pvIfData)
2611	{
2612	/* Nothing to do */
2613	NOREF(pThis); NOREF(pvIfData);
2614	return VINF_SUCCESS;
2615	}
2616

Note: See TracBrowser for help on using the repository browser.

source: vbox/trunk/src/VBox/HostDrivers/VBoxNetFlt/linux/VBoxNetFlt-linux.c@ 104760

Download in other formats: